HCIE-DC V1.0 Training Material 3 Cloud Data Center Computing and Storage Solutions

The privilege of HCNA/HCNP/HCIE:
With any Huawei Career Certification, you have the privilege on http://learning.huawei.com/en to enjoy:
n
 1、e-Learning Courses： Logon http://learning.huawei.com/en and enter Huawei Training/e-Learning
/e

o m
If you have the HCNA/HCNP certificate：You can access Huawei Career Certification and Basic Technology e-Learning
courses.
e i .c
If you have the HCIE certificate: You can access all the e-Learning courses which marked for HCIE Certification Users.
aw

Methods to get the HCIE e-Learning privilege : Please associate HCIE certificate information with your Huawei account, and
hu

email the account to [email protected] to apply for HCIE e-Learning privilege.

g .
 2、 Training Material Download
i n

arn
Content: Huawei product training material and Huawei career certification training material.

//le
Method：Logon http://learning.huawei.com/en and enter Huawei Training/Classroom Training ,then you can download
training material in the specific training introduction page.
p :
 3、 Priority to participate in Huawei Online Open Class (LVC)
t t

s :h
The Huawei career certification training and product training covering all ICT technical domains like R&S, UC&C, Security,
4、Learning Tools: rc e
Storage and so on, which are conducted by Huawei professional instructors.
u

s o
eNSP ：Simulate single Router&Switch device and large network.

R e
WLAN Planner ：Network planning tools for WLAN AP products.

n g
In addition, Huawei has built up Huawei Technical Forum which allows candidates to discuss technical issues with Huawei experts ,
ni
share exam experiences with others or be acquainted with Huawei Products.
a r
 Statement:
L e
r e
This material is for personal use only, and can not be used by any individual or organization for any commercial purposes.
o
M
HUAWEI TECHNOLOGIES CO., LTD. Huawei Confidential 1
Huawei Converged
Resource Pool Solution
n
/e
o m
www.huawei.com
e i.c
w
hua
g.
ni n
Copyright © 2017 Huawei Technologies Co., Ltd. All rights reserved.
ar
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Foreword
 The converged resource pool solution is a cloud data center
solution that is compatible with third-party products and
resources and can better adapt to actual application
n
/e
environments of enterprises. This document describes the
integrated data center solution.
o m
ei.c
w
hua
g.
n i n
ar Page 2
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Objectives
 Upon completion of this course, you will be able to:
 Understand Huawei converged resource pool solution.
 Master design of modules of Huawei converged resource pool

n
/e
solution.
o m
ei.c
w
hua
g.
n i n
ar Page 3
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Contents
1. Overview of Huawei Converged Resource Pool Solution
2. VDC Service
3. Unified O&M Management

n
4. Security Isolation /e
o m
ei.c
w
hua
g.
n i n
ar Page 4
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Converged Resource Pool Solution
Allocate multiple VDCs on demand
Resource pool
n
/e
Cloud service management platform
layer Service Center Operation Center
Resource FusionSphere Data Center 1
o m
i.c
management
layer FusionManager
Virtualization
Fusion Fusion Fusion VMware
...
w e
Data
ua
layer Compute Storage Network Center N
h
Server Storage Network Security
.
Infrastructure
g
layer
n i n
ar Page 5
l e
: //
ttp
 Business-oriented VDC service: Provide the virtual resource pool service for business or
organizations independently.
: h
s
ce
 Unified resource pool management
r
ou
 Unified management for multiple resource pools
es
Unified management for heterogeneous resource pools
R
Unified management for multiple regions and centers
g

ni n
Unified operation and maintenance platform
a r
Unified operation platform, resource-as-a-service, service catalog, and service
L e application approval process
e
or
 Unified resource pool operation and maintenance and rights- and domain-based
management
M  Network security and isolation: The network security and isolation scheme ensures service
security.
Architecture of the Converged Resource
Pool Solution
SC Private cloud resource OC
provisioning
Alarm
performance
n
Project
customization
FM
FM
/e
m
eSight
o
e i.c
w
rd
3 midrange computer
VRM vCenter
ua
management system
Provisioning
. h
Midrange
computer
Server IPSAN Switch Router
i n g
UDS Firewall
r n
l e a Page 6
: //
ttp
 Self-developed ManageOne parts:
: h
Service Center: Resource service center of the solution. It provides simple operation
s

ce
management functions for the enterprise private cloud scenario, including user
r
management, service catalog, service application, approval and release, and resource
self-service management.
s ou

Re
Operation Center: Operation and maintenance center of the solution. It performs
g
operation and maintenance based on scenarios for data center business, provides
n
ni
visual analysis on status, risks, and efficiency, and provides proactive foresight based
a r
on the analysis capability. The operation and maintenance subsystem is divided into
L e the data center operation and maintenance layer (OC) and professional equipment
e
operation and maintenance layer (eSight). The OC is also the unified management
or
portal entry of the solution. In addition to links to OC function, the home page provides
M
unified entries to other part management UIs and supports jump to the links.
 Internal cooperative parts
 Self-develop self-service capability and charging capability provided by the CSB based
on public cloud to end users so that end users can apply for cloud and non-cloud
services
Contents
2. VDC Service

n
o m
ei.c
w
hua
g.
n i n
ar Page 7
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
VDC Service
 Service description:
 A virtual data center (VDC) is built through virtualization technology on the basis of
physical data center. Compared with physical data centers, VDCs also have computing,
storage, and network resources except the resources are virtual.
n
/e
 A VDC is often used to integrate physical resources of a physical data center into a
m
pool, divide the pool into blocks, and provide the blocks to users. VDCs of different
users are mutually isolated and do not interfere with each other.
o
 A VDC has its own manager (or managers). Managers create various resources within
e i.c
w
the resource quota. In addition, other users can apply for resources in a VDC.
ua
 Customer value:

. h
Users are provided with the VDC capability. Resources in a physical data center are
divided into blocks and allocated to users or organizations.
i n g
r n
l e a Page 8
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
VDC Service Model
VDC model
Quota User Resource
Region Shenzhen, Beijing Manager Tom Cloud host WebInstance
CPU 100 Service user Jack Cloud host DBInstance
n
/e
Memory 200 GB Service user Jenney Cloud disk DBDisk
... ... ... ... ... ...
Service Catalog Network Template
o m
i.c
VDC Application template
Cloud host VM template
Cloud disk
MySQL
w e
ua
...
. h
i n g
r n
l e a Page 9
: //
ttp
 A complete VDC includes quota, users, service catalog, network, resources, and templates.

: h
Quota: Quota is used to restrict the upper limit of resources that a VDC can use. The VDC
s
ce
administrator applies to the domain administrator for the VDC quota.
r
ou
 User: A VDC includes users of two types: VDC administrators and service users.
es
Service catalog: The service catalog displays the list of services that VDC users can apply for.
R
The VDC administrator can define the service catalog based on special requirements of the
g
department or organization.
n

r ni
Network: The VDC network includes subnets, IP address segments, security groups, firewall,
e a
and VPNs. It is responsible for communication inside the VDC and for communication and
e L
security protection between the VDC and external networks. The VDC administrator can plan
or
the VDC network topology.
M  Resource: Service resources users have applied for include cloud hosts, cloud disks, and
applications. Users can maintain and monitor the resources.
 Template: VDC users can use global templates and create VDC private application templates
and VM templates.
Shared Resources Implemented by VDC
for Multiple Organizations and Services
XXX_VDC XXX_VDC XXX_VDC
n
/e
OA and Mail BOSS SSDM
m
Gold resources Silver resources Copper resources
o
Resource pool Resource
e i.c
w
1 VMware pool 2
Physical resource Huawei virtual
ua
virtual Physical resource Huawei virtual VMware virtual
pool resource pool pool resource pool resource pool
resource pool
Data Center 1 Data Center 2
. h
i n g
r n
l e a Page 10
: //
ttp
 VDC: After resources are integrated in a pool, a logical data center is flexibly built, which
provides an efficient O&M service platform.
: h
s
ce
 Key capability features:
r
ou
 VDC resource: cloud and non-cloud, heterogeneous cloud, and cross-region
es
VDC independent resource management: The SLA can be guaranteed.
R
Minute-level VDC building and provisioning
g

ni n
VDC self-maintenance
a r
VDC self-operation: self-service provisioning, measurable

L e
VDC service management
e
or  VDC resource management and allocation; service definition; network and security
M policy management; O&M management; user management

Diversified Cloud Services Supported by
VDC
Computing Storage Network
service service service
VDC user

Service application,
use, and management
n
/e
VDC administrator
m

VDC resource
o
management and
i.c
allocation
e
Unified resource pool 
VDC service definition
Computing pool Storage pool Network pool
w
and operation
ua

VDC network and
security policy
h
Resource integration: unified logical resource
.
pool for multiple data centers
management
g
Physical data center 
Data Center 1 VDC O&M management
Data Center 2 Data Center 3
n i n 
VDC user management

ar Page 11
l e
: //
ttp
 Computing service
VM service
: h
s

 X86 physical server

r ce
ou
 Storage service
EBS
es
R

 Network service
n g

r ni
Elastic IP address
e

a SNAT/DNAT
e L ACL
or

 VPN
M
VPC Service
 Service description:
 Virtual private cloud (VPC) is used to help users create a private application operation
environment in cloud through virtualization. After applying for VPC, a user can configure subnets
in the VPC and can configure services such as VFW, VLB, and VM in the VPC.
 The network in a VPC is an isolated environment. Users can plan their private network addresses.
n
Addresses among different VPCs can be overlapped. If an IP address segment is specified for a
/e
m
subnet, the system automatically allocates IP addresses in DHCP mode to VMs in the segment.
o
i.c
 The VPC provides the direct-connect network, routing network, and internal network to meet
requirements of different customers for virtual networking.
 Customer value:
w e
ua
 The virtual private cloud environment in cloud provides users both network isolation and
h
independent maintenance.
g.
n i n
ar Page 12
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
VDC/VPC Network
Department A
V V V
M M M
n
VDC Employees on
/e
VPC2 a business trip
m
V V V
o
M M M
VPC1
Department B
e i.c
w
ua
Internet
. h
i n g
r n
l e a Page 13
: //
ttp
 Isolated environment: The VPC provides isolated VMs and network environments to meet
network isolation requirements of different departments.
: h
s
ce
 Rich services: Each VPC can provide services such as independent firewall, elastic IP address,
r
VLB, security group, VLAN, IPSec VPN, and NAT.

s ou
Flexible networking: three networking modes: direct-connect network, routing network, and
internal network.
Re
n g
r ni
e a
e L
or
M
Contents
2. VDC Service

n
o m
ei.c
w
hua
g.
n i n
ar Page 14
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Resource Pool Expansion and Unified
Management for Multiple Resource Pools
Interconnection between
existing resource pools Unified
management Unified, shared, and
New resource pools are interconnected with existing differentiated
Unified resource management, unified service entry
resource pools. Services are expanded in new
Resources and services are provided through the service
resource pools. Security isolation and network
catalog.
interworking are provided. Unified
management
n
platform
/e
Virtualization
o m
i.c
New resource pools
Server are shared among
e
Value-added
services MSS domain OSS domain BSS domain multiple
Virtualized Virtualized Virtualized Virtualized Physical
Storage departments.
w
resource pool resource pool resource pool resource pool resource pool
ua
New integrated virtualized resource
(Equipment (Equipment (Equipment New pools bear service systems of
room 1) room 2) room 3) resource pool multiple departments.
Across regions and Support physical

. h
g
equipment rooms resource pools.
i n
Support resource pools of multiple Support unified management for physical resource
pools and virtual resource pools.
n
regions and equipment rooms.

ar Page 15
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Unified Management for Heterogeneous
Virtualization
Tenant
Service management
management
Virtual Virtual Virtual load Elastic IP User

VM Virtual disk VDC
network router balancer address management
n
Resource pooling Template management
/e
Computing Storage resource VLAN resource IP address Virtual firewall Virtual load Public network 2
resource pool pool pool resource pool balancer IP address pool VM template
1 1 3
o m
i.c
Resource access layer
Huawei
API
VMware Server Firewall Switch Storage
w e Load
ua
Fusion balancer
vSphere
Compute
1. Obtain information about
. h
2. Create templates on vSphere
g
servers, clusters, and storage 3. Set VLAN
devices from vSphere over API information on the or find existing templates of
n
DVS of vSphere. vSphere.
i
and manage resources in a
resource pool.
r n
l e a Page 16
: //
ttp
 The converged resource pool management platform manages vSphere by invoking APIs
h
issued on the official website of VMware through drivers so as to provide users with the
:
undifferentiated virtual data center service.
s
r ce
Through unified management for virtualization vendors based on an converged resource
ou

pool, end users will not perceive underlying virtualization differences when they use the VDC
service.
es
R

n g
Basic functions such as network management, template management, and VM lifecycle
ni
management (creating a VM, modifying a VM, starting a VM, stopping a VM, logging to a
r
VNC, and deleting a VM) are supported.
a
L e
e
or
M
Unified Management for Heterogeneous
Devices and Platforms
X86 servers of VMware,
vendors FusionCompute
Converged
architecture Physical server
Virtual platform
n
SuSE Linux
/e
Red Hat Linux Converged MS Exchange
CentOS Application
architecture MS SharePoint
Windows
m
WinServer Database
ManageOne
o
OS
i.c
Oracle
MS SQL Server
Huawei
EMC
Storage
Security
Network
w e
ua
NetApp Cisco, Huawei, Juniper
IBM
. h
automatic deployment, and establish a unified IT service environment.
i n g
Standardize the IT software environment based service requirements, increase the service efficiency through
r n
l e a Page 17
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Unified Operation Platform - Resource-as-
a-Service and Lifecycle Management
Product Business
application approval
Self-service Comprehensive
User monitoring for
one-click service
activation
Business
department
n
/e
Service and
product
Converged resource
m
definition
pool
o
Service manager
i.c
Operation center
User
w e
ua
End of service
Service Operation management Resource
template platform recovery
h
definition
.
System
administrator
n g
Provide end-to-end business operation processes and management functions and quickly respond to
markets and users.
i
r n
l e a Page 18
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Unified Operation Platform - VDC Two-
layer Service Catalog
Global service catalog
Cloud host VDC Cloud disk Elastic IP address
n
/e
VDC service catalog of the Development Dept.
o m
Cloud host VDC Cloud disk Elastic IP address
e i.c
1. The VDC service catalog is used by VDC users to apply for services.
w
ua
2. The global service catalog allows all VDCs to be used by default (after a white list is configured, only
specific VDCs can be accessed).
. h
3. VDC administrators can define services based on special requirements of the department or
organization.
i n g
r n
l e a Page 19
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Unified O&M Platform
O&M O&M O&M
manager supervisor personnel
Domain-level O&M Management
Global Trouble ticket event
Operation Center
management
Service Center ITSM
layer Unified alarm
Unified Unified Intelligent
Resources and monitoring topology analysis
n
customer
/e
relations
Cloud
m
resources Non-cloud
DC1n DC2n Monitoring resources
o
Cloud resources information DC3n Monitoring
Non-cloud
i.c
DC12 Monitoring DC22 Non-cloud information
resources DC32
DC11 information Monitoring Cloud resources resources
information Monitoring DC21 Monitoring DC31
e
information information
OpenStack resource pool
DC VRM resource pool Third-party
3rd device management

Neutron resource pool
Fusion Manager
w
management Keystone Fusion
eSight
ua
system
layer
eSight
Ceilometer Manager
3rd Party
Nova Cinda Resource
Fusion
Computer management
h
Glance (VRM)
Heat
Infrastructure Storage Load
g. Middleware &
n
Hypervisor Server Switch Firewall Router OS&DB
application
i
layer device balancer
r n
l e a Page 20
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Unified O&M Platform - Quick Fault
Locating
Search for
resource
instances
Analyze
Solve
recent
problems
alarms
n
Receive fault
reports from
/e
m
customers
o
i.c
Analyze
Rectify faults performance
e
exceptions
w
ua
Analyze
h
Analyze root status of
.
causes associated
resources
i n g
r n
l e a Page 21
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Unified O&M Platform - Multi-dimensional
Presentation of Resource Pool Alarms
Multiple angles of view are provided to view alarms so as to
meet differentiated O&M monitoring appeals.
Service group
n
/e
Customer Service
o m
i.c
User-defined Alarm Management
Dimension system
View
w e
Management system, cluster, and resource pool
ua
Object Object
h
classification Location
Servers, storage devices, network devices...

g.
Data centers, equipment rooms…
n i n
ar Page 22
l e
: //
ttp
 View alarms based on user-defined dimensions: View alarms based on compound conditions
(alarm severity, object, location, and customer).
: h
s
ce
 Support right- and domain-based viewing of alarms.
r
ou
 Support viewing active alarms, alarms being processed, masked alarms, and historical alarms.
es
R
n g
r ni
e a
e L
or
M
Contents
2. VDC Service

n
o m
ei.c
w
hua
g.
n i n
ar Page 23
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Isolation Design of the Converged
Resource Pool Service
Security isolation Security isolation Security isolation
inside services between services between VDCs
Application
DB Service 2 Service 3
Portal server Service 1
Service
manager
n
/e
VPC1
VDC VLAN 1 VLAN2 VLAN 3 VDC 1 VPC2 VDC2
m
manager
o
i.c
Resource
e
pool Unified resource pool
manager
w
ua
DC Resource pool 1
Resource
h
manager DMZ
.
APP Core pool 2
network and increase the IT response speed.

i n g
Through network automation, meet the appeals of service isolation and service deployment for the
r n
l e a Page 24
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Isolation Design of the Converged
Resource Pool Service
Internet
Elastic IP address: 112.70.70.10
VPC
Load balancer
n
/e
Private IP address: Security group
192.160.0.3
PHP PHP Private IP address:

192.160.0.2
o m
i.c
Cloud host
Routing network
Security group
w e
ua
Synchronization
MySQL MySQL
. h
Internal network Cloud host
i n g
r n
l e a Page 25
: //
ttp
 Build a VDC network featuring security isolation by using internal network, routing network,
security group, and ACL.
: h
s
ce
 Deploy the cloud hosts that do not need to interact with outside on the Internal
r
network. External networks cannot perform IP addressing for the internal network. In
ou
the networking diagram, because external networks do not need to access the MySQL
s
e
database, the MySQL cloud host can be placed in the internal network.
R

n g
Deploy the cloud hosts that need to interact with outside on a routing network and
ni
map them to the private IP address of the routing network through an elastic IP
a r
address. External networks can only perform addressing for the elastic IP address. In
L e the networking diagram, an elastic IP address is bound with the load balancer to
e
distribute messages to the private IP addresses of cloud hosts on the routing network.
or
External networks can only perform addressing for elastic IP addresses.
M  Through a security group, control inflow and outflow of information about each cloud
host. In the networking diagram, enable only HTTP ports for PHP cloud hosts and
enable only database access ports for the MySQL database through setting.
 The IP addresses of the internal network and the routing network can overlap between
different VPCs.
 VDC isolation
 VDCs cannot directly access each other, which means different VDCs are isolated by
default (routes are isolated through VRF/vFW and Layer 2 is isolated through
VXLAN/VLAN).
 Different VDCs use different bearer network VPNs and cannot access each other.
 Mutual access between VDCs: VDCs can access each other through only a public
network, which means that the access can be implemented only by applying for
n
EIP/SNAT/DNAT methods for VMs through vFW.
/e
 VPC isolation
o m

e i.c
By default, VPCs cannot access each other, which means different VPCs are isolated by
default.
w

h ua
(Routes are isolated through VRF/vFW and Layer 2 is isolated through VXLAN/VLAN.)
 Subnet isolation
g.
n
Different subnets are isolated through Layer 2 of VXLAN. i n
r

l e a
Users can define the NetAcl policy between subnets and define mutual access rules.
 Internal and external network isolation

: //

h ttp
External networks directly access the internal network: External networks can access
:
VMs on the internal network only when users have applied for public network access
s
ce
(EIP/SNAT/DNAT).
r
ou
 VPN access: Users are allowed to access VMs on the internal network through
s
proprietary data centers only when users have applied for IPSec VPN.
e

R
Users can define the FwAcl policy to implement isolation between internal networks
n g
and external networks.
r ni
The access between internal networks and external networks of vFW is denied by
e a default.
e L
or
M
Quiz
1. vFW is implemented by what physical device and how?
2. What is the value of unified management and maintenance?
n
/e
o m
ei.c
w
hua
g.
n i n
ar Page 27
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Summary
 Services provided in a VDC
 Services provided in a VPC
Characteristics of the Converged Resource Pool Solution's unified

n

management and management

/e
o m
ei.c
w
hua
g.
n i n
ar Page 28
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Thank You
Thank You n
www.huawei.com /e
o m
ei.c
w
hua
g.
ni n
ar
l e
: //
http
s :
r ce
sou
Re
n g
rni
e a
e L
or
M
Converged Resource Pool
Solution - Introduction to
Service Capability Features
n
/e
o m
www.huawei.com
e i.c
w
hua
g.
ni n
ar
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Foreword
 The converged resource pool solution for data centers
implements Huawei cloud data center. By understanding and
mastering features of the converged resource pool service, you
n
/e
can deeply understand Huawei cloud data center solution.
o m
ei.c
w
hua
g.
n i n
ar Page 2
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Objectives
 Master features of the VDC service.
 Understand features of the ECS and physical host services.

n
 Understand and master features of the EVS.
/e
 Understand and master features of the application and network
o m
services.
ei.c
w
hua
g.
n i n
ar Page 3
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Contents
1. Features of the VDC Service
2. Features of the ECS
3. Features of the Physical Host Service

n
4. Features of the EVS /e
o m
i.c
5. Application Service Features
6. Network Service Features

w e
hua
g.
n i n
ar Page 4
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Features of the VDC Service
 Virtual data center (VDC) is a logical data center that flexibly allocates physical resources in a pool
based on service requirements through the logical isolation technology, including the computing,
storage, and network resources (implementing the DCaaS service) required by a data center. It
provides end users with a virtual what you see is what you get (WYSIWYG) data center. The features of
the VDC service are as follows:
n
/e
 VDCs can flexibly match the organizational structure: VDCs can be flexibly classified based on scenario
requirements.
 VDCs can be classified based on departments. Each department can independently manage resources of the
o m
i.c
department and VDCs are mutually isolated.
VDCs can be classified based on fields of use, for example, development VDC and test VDC.
e

VDC resource quota: The resource quota (including CPU, memory, storage, and network resources) can be set for
w

ua
each VDC. The quota of a VDC is measurable and the cost is visual.
 VDC resource building supports multiple physical data centers and unified resource allocation.
. h
Self operation and maintenance: Each VDC is an independent management entity that provides the self operation
i n g
and maintenance capabilities. In a VDC, users can apply for, manage, and monitor IT resources.
r n
l e a Page 5
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
VDC Model Diagram
Domain
DC-1 DC-2
AZ-1 AZ-2 AZ-3 AZ-4
n
VPC-1 VPC-2 VPC-3 VPC-4 VPC-5 VPC-6
/e
Host SLA
SLA SLA SLA
aggregate -1
Host aggregate -1 Host aggregate -2
m
Host aggregate -2
d d d
o
SLA SLA SLA
i.c
SLA Host Host
Host Host aggregate -3 aggregate -4
e
aggregate -3 aggregate -4
VDC-2 VDC-3
VDC-1
w
h ua
g.
n i n
ar Page 6
l e
: //
ttp
 Domain: A domain involves the scope of global resource management and provides the
h
capability of accessing multiple data centers (DCs) and multiple resources.
:
s
ce
 DC: A DC refers to the collection of computing, storage, and network resources with the same
r
network egress. It may include multiple available zones (AZs).

s ou
AZ: An AZ refers to a zone for physical resources (computing, storage, and network
Re
resources). In an AZ, Layer 2 of the physical network supports interworking. Users can bind
g
virtual machines (VMs), disks, and networks in an AZ without limitations.
n
ni
 Host aggregate: Host aggregate refers to the collection of computing resources that share the
a r
network and storage resources and have the same SLA. It is similar to a cluster.

L e
Virtual private cloud (VPC): A VPC creates an isolated network environment for application
e
or
systems. It helps users completely control their virtual networks, including selecting the IP
address range, allocating network segments, and configuring a routing table and gateway.
M
VDC Application
 VDC application supports the following modes:
 A domain service manager creates VDCs and allocates the VDCs to VDC service
administrators.
n
 VDC service administrators apply for VDCs through the service catalog and domain
/e
service managers approve the applications and distribute VDCs.
 Customer value:
o m
i.c
 Quick resource allocation: Resources are quickly allocated as required to increase the
efficiency of the general resource allocation process.
w e
ua
 Flexible resource application modes are provided: Domain service manager can
directly allocate VDCs and VDC service administrators can apply for VDCs
independently.
. h
i n g
r n
l e a Page 7
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Entry of VDC Application
 Domain service managers directly create VDCs through the manager view.
 VDC service administrators apply for VDCs through the service catalog of the
tenant view.
Service
n
/e
Service
Portal Template Application
catalog management
o m
ei.c
w
VDC of the Development Dept VDC of the Testing Dept
hua
.
Shenzhen Data Shanghai Data Center Beijing Data Center
g
Center
n i n
ar Page 8
l e
: //
ttp
 Name: unique name of a VDC.

: h
Expiration date: date on which the VDC service expires (the date is entered by a VDC service
s
ce
administrator during application for a VDC or is specified by a domain service manager during
approval).
r

s ou
VDC quota: upper limit of resources that can be allocated to a VDC (VDQ quota is entered by
Re
the VDC service administrator during application for a VDC or is specified by a domain service
g
manager during approval).
n
ni
 AZ: A domain service manager can plan AZs for a VDC. When a VDC service administrator
a r
automatically applies for VDCs, the VDC service administrator cannot specify this parameter.
L e
This parameter is controlled by the domain service manager.
e
or
 VDC member: Specifies VDC administrators and service users in a VDC. This parameter is
specified only when a domain service manager allocates VDCs. When a VDC service
M administrator applies for a VDC, the VDC service administrator cannot specify this parameter.
VDC Change
 Description:
 Based on requirements for use of VDC resources, a VDC service
administrator can apply to a domain service manager for changing
n
/e
the VDC resource quota.
 When existing AZs of a VDC cannot meet service use requirements,

o m
i.c
a domain service manager can add AZs for the VDC.
 When VDC organizational members are changed, a VDC service

w e
administrator can change VDC members.
A domain service manager can change the name of a VDC. u

a

. h
n g
r ni
l e a
Copyright © 2017 Huawei Technologies Co., Ltd. All rights reserved. Page 9
: //

t tp
After a VDC is created, VDC members, AZs, and resource quota can be changed.
 Customer value:
: h
e s

r c
Business development requirements of customers are met and dynamic resource
expansion is supported.
o u

e s
Organizational change requirements of customers are matched.
R
in g
r n
e a
e L
or
M
VDC Release
 When VDCs are no longer used, a VDC service administrator
requests a domain service manager to release the VDCs. In this
case, the system automatically recovers the used cloud resources.
n
/e
 Customer value:
 Resources can be used repeatedly and the resource cost is reduced.
o m
 When resources are released, data on disks used by services can be
ei.c
permanently erased to ensure security of service data.
w
hua
g.
n i n
ar Page 10
l e
: //
ttp
 A domain service manager is responsible for releasing VDCs.
: h
s
r ce
s ou
Re
n g
r ni
e a
e L
or
M
VDC Service Catalog Management (1)
 VDC service catalog management includes:
 Service catalog management: The service catalog is used to store
services with the same attributes.
n
/e
 Service management: Creating, modifying, publishing, unpublishing,
and deleting a service in a VDC are supported.
o m
i.c
 Customer value:
 Services are classified through the service catalog to facilitate
w e
management and maintenance of managers.
hua
 Flexible and customizable service operation processes are provided.
g.
n i n
ar Page 11
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
VDC Service Catalog Management (2)
 1. The VDC service catalog is used by service users to apply for services.
 The VDC service catalog inherits the global service catalog. VDC service
administrators can define services based on special requirements of the
n
department or organization.
Global service catalog
/e
o m
i.c
ECS VDC Cloud disk Elastic IP address
w e
ua
VDC user-
defined
VDC service catalog of the Development Dept services
. h
ECS VDC Cloud disk Elastic IP Compilation
i n g
n
address environment

ar Page 12
l e
: //
ttp
 A domain service manager provides the following global service catalog management
functions:
: h
s
ce
 Creating, modifying, and deleting the catalog.
r
ou
 Creating services (VDC, ECS, cloud disk, and elastic IP address).
es
Setting service permissions (VDCs that can be used).
R
Publishing a service.
g

ni n
Unpublishing a service.
a r
Editing a service.
L e Deleting a service.
e

or  VDC service administrators provide the following service catalog management functions for
M services in a VDC:
 Creating, modifying, and deleting the catalog.
 Creating an application service.
 Publishing an application service.
 Unpublishing an application service.
 Editing an application service.
 Deleting an application service.

VDC User Management
 Customer value
 Customers are provided with flexible permission management and VDC service
administrators can manage users in VDCs.
n
 A VDC provides multiple VDC service administrators to increase the operation and
/e
maintenance efficiency.
 A VDC involves two user roles:
o m
i.c
 VDC service administrator: VDC service administrators are responsible for applying,
monitoring, maintaining, and managing VDCs.
w e
ua
 Service user: Service users are consumers of VDC resources and can apply for services
from the VDC service catalog.
. h
i n g
r n
l e a Page 13
: //
ttp
 The function of managing users and roles in a VDC is provided.

: h
VDC service administrators can manage VDC service administrators (at the same level) and
s
ce
VDC service users and provide the following management functions:
r
ou
 Creating a user.
 Modifying a user
es
Deleting a user. R
g

ni n
Querying a user.
a r
Limiting the user login time and IP addresses of user clients.
L e Resetting a user password.

e

or  Locking or unlocking a user.
M  Service user
 Applying for a service.
 Resetting the password of a user.
 Managing service instances a user has applied for.
 Restriction of use
 The current version supports only two roles, namely VDC service administrator and
VDC user, and does not support user-defined roles.
 One VDC service administrator can apply for one VDC only.
VDC Self-Maintenance
 A VDC service administrator can complete the following self-maintenance work through
the tenant portal:
 Query the usage, total amount, used amount, and unused amount of VDC quota indicators.
 Collect statistics of applied resources based on resource types (ECS, cloud disk, and physical
machine).
n
 Provide monitoring for performance of resources in a VDC, including CPU utilization, memory
/e
usage, network inflow rate, disk write speed, disk read speed, and disk usage.
o m
i.c
 Configuring the VDC network environment, including creating a VPC, subnet, security group,
e
virtual firewall, virtual router, and elastic IP address.
 Service user:
w
ua
 Managing the life cycle of services a user has applied for, including starting, restarting, and
. h
shutting down a ECS, loading and unloading a cloud disk, starting and stopping an application
instance, and binding and unbinding an elastic IP address.
i n g
r n
l e a Page 14
: //
ttp
 VDC service administrators and service users can manage resources in VDCs and networks
and monitor the resource capacity and usage.
: h
s
ce
 Virtual firewall depends on hardware firewall.
r
ou
 Virtual routers depend on hardware routers and switches.
es
The following hardware firewall models for virtual firewall are supported: Eudemon
R
E1000E-N and Eudemon E8000E-X.
n g
ni
 The following switch model for virtual routers is supported: CloudEngine 12800.
a r
L e
e
or
M
VDC Resource Quota Measurement
 The VDC quota measurement function provides measurement data for service settlement.
Managers and tenants perform charging offline based on measurement data. At present,
the supported VDC measurement quota indicators include: number of vCPUs, memory,
disk, elastic IP address, number of VPCs, number of VMs, and number of security groups.
n
/e
 VDC quota measurement includes the following functions: querying statistical
measurement data of a VDC in the specified time segment and exporting statistical
measurement data.
o m
Create/modify/del Trigger/terminate
e i.c
w
ete a VDC. measurement
ua
and dotting.
Domain administrator SC-Meter
Apply for/change/
h
delete a VDC. Periodic dotting
VDC administrator
g.
SC–Meter database
n i n
ar Page 15
l e
: //
ttp
 Measure quota of resources in a VDC, for example, CPU, memory, disk, VPC, elastic IP
address, security group, and VM.
: h
s
ce
 The system measures resources whose quota is limited in a VDC every hour and collects
r
statistics of accumulated measurement results by default.

s ou
VDC measurement supports export based on time segment. Export in only the Excel format is
supported.
Re

n g
Customer value: VDC quota measurement data is provided for domain service managers and
ni
VDC service administrators to make settlement offline.
a r
L e
e
or
M
Contents

n
o m
i.c

w e
hua
g.
n i n
ar Page 16
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
ECS
 Application for a ECS:
 Service users can apply for the specified ECS from the VDC service catalog.
 Users can perform the following operations on the ECSs users have applied for:
 Loading or unloading a cloud disk on a ECS.
n
 Adding or deleting a network interface card.
/e
 Power-on, power-off, restart, and hibernation
o m
 Changing specifications of a ECS (CPU, memory, disk, and number of network
interface cards).
ei.c
w
ua
 Changing the expiration date of the ECS.
 Releasing a ECS.
. h
i n g
r n
l e a Page 17
: //
ttp
 The ECS provides users with the VM service that can be used after the service is applied for.
h
Users can flexibly apply for ECSs of the specified CPU, memory, disk, network interface card
:
s
specification, or OS type to meet computing requirements of various applications.
Customer value
r ce
ou

es
Resources are dynamically allocated and shared.

R
Different requirements of end users are met and quick provisioning is provided.
n g
ni
 The user-friendly self-service function increases the work efficiency and reduces the
r
O&M cost.
e a
Restriction of use
L

e
or
 Creating ECS snapshots, restoring a ECS from snapshots, and deleting snapshots are
not supported.
M  Cloning a ECS or converting a ECS into a VM template is not supported.
 Managing VMs on the existing virtualized platform is not supported.

Contents

n
o m
i.c

w e
hua
g.
n i n
ar Page 18
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Physical Machine Service
 Application for the physical machine service:
 Region
 AZ
 Specification
n
/e
 OS
m
 User operation:
Users can remotely log in to a physical machine (in non-BMC mode) and deploy required
o
i.c

applications.
 When the physical machine service expires, users can apply for extension of the service.
w e
ua
 When a physical machine is not used, users can apply for releasing the physical machine service.
When a physical machine is released, the cloud management platform clears only service
. h
instances and data related to users. Data on the physical machine needs to be cleared by the
administrator.
i n g
r n
l e a Page 19
: //
ttp
 The physical machine service is generally used in applications where ECSs are inapplicable,
h
for example, some large data analysis services are directly deployed on physical machines to
:
obtain lower network I/O delay than that of ECSs.
s
r ce
VDC service users apply for physical machines. After VDC service administrators approve the
ou

applications, the system automatically deploys the physical machine OS and provides it to
service users.
es
R

n g
AZ: In terms of design, AZs are mutually independent, that is, AZs have their independent
ni
power supply and independent physical network devices. When an AZ fails, other AZs are not
r
affected. For example, the active and standby ECSs of an application or ECSs supporting load
a
e
balancing are deployed in different AZs to improve reliability.
L
e
or
The feature is a test feature and strictly restricted.

M

 Restriction on automatic deployment of physical machines in the current version:
 The PXE network and service network need to be configured manually and only
VLAN is supported.
 Only Huawei servers RH2288, RH2285, and RH5885 are supported.
 Automatic deployment of only the SUSE11SP3 and SUSE11SP2 OSs is supported.

 OSs can only be installed on local disks of physical servers.
 Users cannot power on or power off physical machines or log in to physical

machines in BMC mode.
n
/e
o m
e i.c
w
h ua
g.
n i n
ar
l e
: //
h ttp
s :
r ce
s ou
Re
n g
rni
e a
e L
or
M
Contents

n
o m
i.c

w e
hua
g.
n i n
ar Page 21
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
EVS
 The EVS is combined with ECSs to provide the persistent block storage service.
 Cloud disks can be mounted under ECSs in one AZ or unloaded from ECSs
independent of the life cycle of ECSs. When a user applies for a ECS, the user
n
can specify the capacity and storage SLA (specified storage medium: SATA, SAS,
SSD, or Any. The storage SLA option is defined by the administrator in the
/e
resource pool).
o m
i.c
 Customer value

w e
Cloud disks provide the persistent high-availability block storage device service.
ua
Customers can quickly complete cloud disk application and flexibly adjust the cloud
disk specifications and storage SLA online based on actual requirements. Cloud disk
. h
instances can be mounted on instances of running ECSs and are used with ECSs. In
this way, data of customers is more secure and flexible.
i n g
r n
l e a Page 22
: //
ttp
 Application for cloud disks: Service users can apply for cloud disks of the specified capacity
and storage SLA from the VDC service catalog.
: h
s
ce
 Region: For the environment that covers multiple data centers across regions, users can apply
r
for cloud disks nearby to reduce delay.

s ou
AZ: In terms of design, AZs are mutually independent, that is, AZs have their independent
Re
power supply and independent physical network devices. When an AZ fails, other AZs are not
g
affected. For example, the active and standby ECSs of an application or ECSs supporting load
n
ni
balancing are deployed in different AZs to improve reliability.

a r
VPC: A VPC provides application systems with isolated network environment to meet
L e
different network isolation requirements. Select the VPC of ECSs that are bound based on
e
or
cloud disk planning.
 Capacity: cloud disk size. Users can define the capacity. The unit is GB.
M  Type: Cloud disks include common and shared cloud disks. Common cloud disks can be
mounted on one ECS and shared cloud disks can be mounted on multiple ECSs at the same
time. (When multiple VMs use one shared disk, if data is written at the same time, data may
be lost. If shared cloud disks are used, the application must ensure access control over the
cloud disks.)
 Storage SLA: The storage SLA option is defined by a resource pool and is discovered by the
cloud management platform. The administrator can set the SLA label on the storage pool, for
example, storage SLA defined based on the RAID level or disk type.
 Sources: mirror, disk, and snapshot.
 Mirror: Create a cloud disk through VM mirror. The created cloud disk includes content
of the VM mirror.
 Disk: Create a disk through an existing disk. The created source disk is consistent with
n
the existing disk in terms of content.
/e
 Snapshot: Create a disk through disk snapshot.
o m
 Number of disks: A user can apply for one or more cloud disks at the same time.
e i.c
w
ua
 Expiration date: date on which the EVS expires.
 Limitation
. h

n g
The bind/unbind operation can only be performed on block storage resources that are
i
n
separately applied. The system disks carried when VMs are applied for do not support
r
the bind/unbind operation.
l e a

: //
The storage space occupied by cloud disk snapshots is not limited by the VDC quota.
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Contents

n
o m
i.c

w e
hua
g.
n i n
ar Page 24
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Application Service (1)
 The application service is the collection of VMs and software that provide a
certain service, for example, applications such as SQL Server, CRM, and ERP.
 Application orchestration:
 Domain service managers and VDC service administrators can complete template
n
design and preparation by manually compiling template scripts online or exporting
/e
templates. Templates can be debugged, modified, exported, published, unpublished,
o m
i.c
or deleted.
 Publishing an application service:
w e
ua
 VDC service administrators can define published application templates as application
services and publish, unpublish, or delete application services.
. h
i n g
r n
l e a Page 25
: //
ttp
 A simple and one-click deployment service that implements applications is provided to
shorten the online time.
: h
s
ce
 Application orchestration: Domain service managers and VDC service administrators can
r
complete template design and preparation by manually compiling template scripts online or
ou
exporting templates. Templates can be debugged, modified, exported, published,
s
Re
unpublished, or deleted. Domain service managers can prepare general application
templates and provide the templates to VDC service administrators. Application templates
n g
define the ECSs, networks, storage resources, scaling policies, dependencies, and scripts
r ni
required by applications.
e a
e L
or
M
 Application deployment legend
1 2 3
Prepare for Select a resource Create a scaling

deployment. pool. group.
n
4 6
7 8
/e
m
Perform initial
o
Install Apache. configuration of
Create VM A.
i.c
Apache.
5
Install Tomcat.
e
Create VM B1. Add VM B1 to the
scaling group.
w
ua
Create VM B2. Add VM B2 to the Install Tomcat.
scaling group.
. h
Create VM C. Install MySQL.
i n g
r n
l e a Page 26
: //
ttp
 Application deployment: Service users apply for application services through the VDC service
h
catalog and deployment applications at one click. After service users select application
:
s
templates and enter parameters defined by applications, the system automatically creates
r ce
VMs, installs applications on the VMs, establishes network connections, creates
ou
dependencies between applications, and completes application deployment.
es
R
n g
r ni
e a
e L
or
M
 Flexible scaling of applications
 During running of an application, the automatic scaling policy in the group
dynamically adjusts the number of ECSs actually used by the application based on the
current load of the application (the flexible scaling policy is specified in the prepared
n
/e
application script).
m
 When the load of an application exceeds the preset threshold (CPU and memory), the
system automatically adds ECSs to the application and installs the application to
o
reduce the overall resource load of the application and make the application run in a
healthy manner.
e i.c
w
ua
 When the load of an application is lower than the threshold, the system automatically
h
reduces the number of VMs used by the application and releases corresponding
g.
resources to achieve effective resource reuse among applications, energy saving, and
emission reduction.
n i n
ar Page 27
l e
: //
ttp
: h
The design of application templates does not support graphical drag.
s

r ce
Application templates do not support vLB capability design.
ou
 The automatic deployment application needs to provide the automatic installation
es
script of an application and preset it in the mirror.
R
Application monitoring is not supported.
g

ni n
a r
L e
e
or
M
Contents

n
o m
i.c

w e
hua
g.
n i n
ar Page 28
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Elastic IP Address Service
 An EIP address is a static external IP address. EIP addresses are associated with ECSs.
Through EIP addresses, users can access VMs from the Internet. When an associated VM
fails or needs to be upgraded, the EIP address can be quickly mapped to another VM that
is working properly without changing the configuration of the VM client so that services
can be continuously obtained from the standby VM and the impact on services can be
n
reduced.
/e
o m
e i.c
w
h ua
g.
n i n
ar Page 29
l e
: //
ttp
 The elastic IP address service reduces the number of public network IP addresses occupied by
customer services and saves costs.
: h
s
ce
 Application for elastic IP addresses:
r
ou
 Region: For the environment that covers multiple data centers across regions, users can apply
s
for elastic IP addresses nearby to reduce delay.
e

R
VPC: Specifies the VPC under which elastic IP addresses are applied for.
n g
ni
 Specification: Specifies the bandwidth of elastic IP addresses and applies for elastic IP
r
addresses from the static external IP address pool. The specified specification is supported
e a
only when VPCs adopt hardware routers.
e L
Number of elastic IP addresses: A user can apply for one or more elastic IP addresses at the
or
same time.
M  Expiration date: date on which the elastic IP address service expires.
 Elastic IP address operation: Users can perform the following operations on the applied
elastic IP addresses:
 Bind an elastic IP address to the network interface card of a ECS. (Directly configure an
elastic IP address on the network interface card.)
 Bind an elastic IP address to a private IP address.

 Unbind an elastic IP address from a ECS or private IP address.
 Change the bandwidth of an elastic IP address.
 Change the expiration date of the elastic IP address service.
 Extend the elastic IP address service.
 Release an elastic IP address.
Restriction of use
n

 This feature depends on virtual firewall and cannot be used when no virtual firewall is
/e
provided.
o m
 Physical firewall equipment needs to be configured manually.
e i.c
w
h ua
g.
n i n
ar
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Quiz
1. What is the relationship between the VDC and VPC?
2. What are the main service components of a VPC?
n
/e
o m
ei.c
w
hua
g.
n i n
ar Page 31
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Summary
 Main service features of a VDC
 Functions and application scenarios of EIP
n
/e
o m
ei.c
w
hua
g.
n i n
ar Page 32
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Thank You n
www.huawei.com /e
o m
ei.c
w
hua
g.
ni n
ar
l e
: //
http
s :
r ce
sou
Re
n g
rni
e a
e L
or
M
Converged Resource Pool Solution
- Introduction to Resource Pool
O&M Capability Features
n
/e
o m
www.huawei.com
e i.c
w
hua
g.
ni n
ar
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Foreword
 The converged resource pool solution for data centers
implements Huawei cloud data center. By understanding and
mastering features of the converged resource pool service, you
n
/e
can deeply understand Huawei cloud data center solution.
o m
ei.c
w
hua
g.
n i n
ar Page 2
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Objectives
 Understand centralized alarm management.
 Understand centralized performance management.

n
 Understand centralized topology management.
/e
 Understand centralized resource O&M.
o m
 Master report management.
ei.c
Master data center health analysis.
w
ua

. h
i n g
r n
l e a Page 3
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Contents
1. Centralized Alarm Management
2. Centralized Performance Management
3. Centralized Topology Management

n
4. Centralized Resource O&M /e
o m
i.c
5. Report Management
6. Data Center Health Analysis

w e
7. Cloud Resource Pool Capacity Management
hua
g.
n i n
ar Page 4
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Multi-dimension Alarm View
 Multiple dimensions and angles of view are provided to browse
alarms.
 Presentation dimensions:
n
 Range of presented alarms
/e
 Search and sorting
o m
 Viewing of alarm details
ei.c
w
hua
g.
n i n
ar Page 5
l e
: //
ttp
 Multiple measures for obtaining information about alarms or events are provided. O&M
h
personnel can select a target dimension to monitor alarms or events.
:
s
ce
 Based on alarm maintenance scenarios, information required by alarm analysis and
r
processing is provided in a centralized way to reduce user operation and help users obtain
necessary information.
s ou
Re

n g
Present alarms or events based on types of monitored objects.
r ni
Present alarms or events based on physical locations.
e a Present alarms or events based on virtual logic.

L

e
or
 Present alarms or events based on the source management system.
Present alarms or events based on the customer dimension.

M

 Present alarms or events based on services.
 Present alarms or events based on user-defined conditions.
 Present alarms or events based on the VDC dimension.
 Range of presented alarms
 Active alarms
 Alarms being processed
 Historical alarms
 Masked alarms
Multi-dimension Alarm View
Multiple dimensions are provided to view alarms.
Range of presented alarms
n

 Search and sorting

/e
Viewing of alarm details
o m
i.c

w e
hua
g.
n i n
ar Page 6
l e
: //
ttp
 View alarm details. Support presentation of the selected alarm status, associated objects,
h
alarms of associated objects, details of alarm objects, and performance status of alarm
:
objects.
s
r ce
Alarm list: Displayed fields include: alarm severity, alarm name, source device name,
ou

monitoring system name, IP address, first alarm time, last alarm time, locating information,
es
possible cause, confirm status, trouble ticket No., attachment information, serial number,
R
alarm ID, source system, frequency, notification status, cleared status, trouble ticket sending
n g
status, masked status, alarm validity, confirm user, clear user, confirm time, dispatch time,
r ni
clear time, subobject, and alarm remarks.

e a
Alarm details, including content of the alarm list. In alarm details, editing alarm remarks and
e L
maintenance experience is supported.
or  Associated object: View topologies associated with the object; support export of resource
M 
information in the Excel format.
Alarm of associated object: View the alarm list of the associated object.
Alarm Masking
 By setting alarm masking, avoid invalid alarms.
 During device expansion and maintenance in a data center, set alarm masking to avoid
invalid alarms.
Alarm masking rules are as follows:
n

/e
 Support setting alarm masking rules based on specific alarm object instances.
m
 Support setting alarm masking rules based on the range of specified alarm objects (for example,
resource category, source system, physical location (logical resource zones such as AZ and VPC),
o
i.c
customer, and service group).
 Support setting alarm masking rules based on the specified alarm severity.
w e
ua
 Support setting alarm masking rules based on the specific alarm type.
Support setting alarm masking rules based on the preceding compound conditions.
h

 Support specifying the time range in which alarm masking rules take effect.
g.
 Support recording names of alarm masking rules.
n i n
ar Page 7
l e
: //
ttp
 Mask invalid alarms in the device engineering period to eliminate interference to normal
h
alarms and focus on critical alarms to increase the maintenance efficiency.
:
s
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Alarm Consolidation
 The consolidation function for repeated alarms is provided, that is, alarms with
the same source are merged into one alarm, and information such as the first
alarm time, last alarm time, and alarm frequency is provided.
n
 The system provides the automatic consolidation function for repeated alarms,
that is, alarms with the same source are merged into one alarm and
/e
information such as the first alarm time, last alarm time, and alarm frequency in
o m
i.c
the information about consolidated alarms is provided.
 Manual alarm merging is also supported.
w e
hua
g.
n i n
ar Page 8
l e
: //
ttp
 After repeated alarms are merged, more valuable information is provided.
: h
s
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Alarm Notification
 After alarm notification rules are set, when alarms meeting conditions are received, users are
automatically notified of the alarms by email, short message service (SMS), or sound.
 Based on preset rules, after alarms meeting the rules are received, the user group is notified by email or SMS.
 After an alarm is cleared, an alarm notification is sent.
n
 Support update of overdue notifications.
/e
 Support defining alarm notifications based on a template.
m
 Support alarm sound. Different sound can be set for different levels.
o
 Support setting alarm notification rules based on specific alarm object instances.
i.c
 Support setting alarm notification rules based on the range of specified alarm objects (for example, resource
e
category, source system, physical location (logical resource zones such as AZ and VPC), customer, and service
w
group).
ua
 Support setting alarm notification rules based on the specified alarm severity.
h
 Support setting alarm notification rules based on the specified alarm type.
 Support setting alarm notification rules based on the preceding compound conditions.
g.
n i n
ar Page 9
l e
: //
ttp
 After alarm notification rules are set, when alarms meeting conditions are received, users are
h
automatically notified of the alarms by email, SMS, or sound. Based on the alarm definition
:
s
and screening functions provided by the system, the angles of view for viewing alarms can be
defined more flexibly.
r ce

s ou
After alarm notification rules are flexibly configured, users can be informed of alarm events
e
they concern about the most so as to take measures.
R

n g
Restriction of use:
ni
 Notification by email requires support of the SMTP mail server.
a r
e
 Notification by EMS requires support of short message gateway interface protocol 1.2
e L (SGIP1.2).
or
M
Conversion of Alarms into Trouble Tickets
 By setting alarm severity and redefining rules, alarm severity of the specified conditions is
automatically adjusted to meet operation and maintenance management requirements.
 By setting conversion of fault alarms into ITIL events (trouble tickets), the fault response efficiency can
be increased.
n
 Based on preset rules, after an alarm meeting the rules is received, a trouble ticket event is automatically created
/e
in the ITIL system.
Support manually creating a trouble ticket event in the ITIL system based on the specified alarm.
m

o
 If an alarm is not handled several minutes later after it is generated, the alarm is automatically converted into a
i.c
trouble ticket.
e
 If an alarm is not generated several times, the alarm is automatically converted into a trouble ticket.
w
 Support setting rules for automatically converting alarms into ITIL events based on specific alarm object instances.
ua
 Support setting rules for automatically converting alarms into ITIL events based on the range of specified alarm
objects (for example, resource category, source system, physical location (logical resource zones such as AZ and
VPC), customer, and service group).
. h
i n g
r n
l e a Page 10
: //
ttp
 Automatic and manual creation of trouble tickets.

: h
Normalize the troubleshooting process and increase the fault response efficiency and
s
ce
proactive maintenance efficiency.
r
ou
 Support setting rules for automatically converting alarms into ITIL events based on the
specified alarm severity.
es

R
Support setting rules for automatically converting alarms into ITIL events based on the
g
specific alarm type.
n

r ni
Support setting rules for automatically converting alarms into ITIL events based on the
e a
preceding compound conditions.
e L
Restriction of use:
or  The current version does not provide the trouble ticket system. If required, purchase a
M trouble ticket system from a third party. In this case, interworking is involved.
Alarm Knowledge Base Management
 The alarm knowledge base is a dynamic experience library about
processing of historical alarms.
 The alarm knowledge base provides administrators with a

n
/e
dynamic experience library about processing of alarms.
 On a new alarm that is generated, you can view the experience in
o m
processing the same alarms.
ei.c
 After processing a fault alarm, you can enter the processing
w
experience in the knowledge base.
hua
g.
n i n
ar Page 11
l e
: //
ttp
 A daily troubleshooting experience sharing mechanism is provided to facilitate O&M skill
accumulation and sharing of O&M personnel.
: h
s
ce
 Processing experience in the alarm knowledge base helps increase the fault locating and
troubleshooting efficiency.
r

s ou
The alarm information definition function can be used to maintain alarm help information
collected in the system.
Re

n g
Defining an alarm name is supported.
r ni
Self-service editing of alarm help information is supported.
e a Batch import of alarm help information based on a template is supported.

L

e
or
M
Contents

n
o m
i.c

w e
hua
g.
n i n
ar Page 12
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Display of Top N Performance Events
 In a data center, sort common performance indicators of all devices and virtual resources
in terms of Top N and sort mean values in terms of Top N so that administrators can
master information about devices and resources with the greatest performance pressure.
 Array storage: read/write IOPS, read/write IO size, IOPS, and CPU utilization.
n
/e
 Router: CPU utilization, memory usage, interface transmit rate, interface receive rate, interface
outflow bandwidth utilization, and interface inflow bandwidth utilization.
 Firewall: CPU utilization, memory usage, interface transmit rate, interface receive rate, interface
o m
i.c
e
 Switch: CPU utilization, memory usage, interface transmit rate, interface receive rate, interface
w
ua
 Server: CPU utilization and memory usage.
. h
Cloud VM: CPU utilization, memory usage, disk usage, network I/O flow rate, and disk I/O
frequency.
i n g
r n
l e a Page 13
: //
ttp
 The Top N performance view about common indicators of a data center is provided.

: h
Users can quickly learn the general health and bottleneck of the data center.
s
r ce
s ou
Re
n g
r ni
e a
e L
or
M
User-defined Performance View
 Users define resources and indicators to generate a user-defined
performance view.
 The user-defined performance view reduces the workload of selecting

indicators and devices every time and facilitates customers to view
n
performance data of devices. /e
o m
i.c
 By specifying the resource category and its performance indicators,
complete definition of a performance view.
w e

u aand
Resources can be classified based on objects (server, storage resource,
h
network device, VM, and cloud disk), physical location, customer,
.
cluster.
n g
r ni
l e a
: //
 Performance indicators include:
t tp
Cloud VM
: h
s

 Data storage
r c e
 Server
o u
Firewall
e s
R

in g
Array storage
r n
Network devices
e a
Real-time performance monitoring:
e L The real-time performance of the specified indicator of the specified object can be
r

o monitored.
M  The current performance status of resources is monitored in real time.
 At present, virtual resources can be monitored based on the granularity of one minute
and other devices are monitored based on the granularity of five minutes.
Viewing of Historical Performance
 The historical performance data of the specified indicator of the
specified object can be viewed.
 The historical performance data in the recent one hour can be displayed.
n
/e
 The historical performance data in the recent one day can be displayed.
m
 The historical performance data in the recent seven days can be displayed.
o
i.c
 The historical performance data in the recent 30 days can be displayed.
 The historical performance data in a user-defined time range can be
w e
ua
displayed.
. h
i n g
r n
l e a Page 15
: //
ttp
 View historical performance data of monitored resources to meet the fault locating
requirement.
: h
s
ce
 Restriction of use:
r
ou
 The current version only supports obtaining resource performance data from eSight
and FusionSphere.
es
R
n g
r ni
e a
e L
or
M
Performance Warning
 The performance threshold setting is supported:
 Support selecting the categories and performance indicators of monitored objects
and setting the alarm threshold. Alarm thresholds and recovery thresholds of
different levels can be set.
n
/e
 Support selecting a monitored object and its performance indicators and setting the
m
alarm threshold. Alarm thresholds and recovery thresholds of different levels can be
set.
o
 Support batch selecting monitored objects and their performance indicators and
e i.c
w
setting the alarm threshold. Alarm thresholds and recovery thresholds of different
ua
levels can be set.
. h
Support defining a threshold based on the upper or lower limit of performance.
i n g
r n
l e a Page 16
: //
ttp
 After a performance alarm threshold is set, when the performance of a monitored object
h
exceeds the threshold, an alarm is automatically generated. This provides a timely warning
:
for performance of devices in the data center.
s
r ce
Means for preventing faults is provided to help users take measures in time to eliminate
ou

potential troubles.
es
R
 Performance statistical report: The system provides the performance statistical report
g
function to collect statistics of historical data by category and provide reports. Reports
n
ni
include performance reports, resource reports, capacity reports, alarm reports, and trouble
r
ticket reports.
a
L e
e
or
M
Contents

n
o m
i.c

w e
hua
g.
n i n
ar Page 17
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Physical Topology Display
 Support querying topologies based on physical locations.
 Learn the system from the physical dimension and search for
resources through navigation.
n
 Support searching for topologies based on physical locations,
/e
specifically, region, data center, zone in a data center, rack, and
o m
device.
ei.c
w
hua
g.
n i n
ar Page 18
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Distributed Topologies of Multiple Data
Centers
 Users can view multiple data centers from topologies, which also
show connections between data centers. The system supports
rights- and domain-based management for multiple data centers
n
/e
and can drill down data center with permissions.
 Topologies of VDCs and physical DCs can be shown.

o m
ei.c
w
hua
g.
n i n
ar Page 19
l e
: //
ttp
 Topologies can be viewed based on the dimension of data center distribution.
 Multiple data centers are maintained in a unified way.

: h
s
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Logical Topology Display
 Topologies can be viewed based on the logical resource dimension.
 Resources are viewed from the logical concept. Virtual resources can be viewed.
 Resources are classified based on logical concept, including physical resources

and virtual resources. When users view topologies from logical objects, users
n
can choose to view topologies based on resource pools and unfold resources in
/e
the resource pool management system or choose to unfold resources based on
o m
i.c
AZ until virtual resources are found.
w e
hua
g.
n i n
ar Page 20
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Service Application Topology
 From the angle of resource allocation and use, resources are used by
application services and can be viewed based on service group. Services often
belong to VDCs of the specified customer. Resources can be unfolded from
VDCs to services and then to resources. Topologies of VAPP and service group
n
/e
applications can be shown.
o m
ei.c
w
hua
g.
n i n
ar Page 21
l e
: //
ttp
 Topologies can be viewed based on the service application dimension.

: h
Helps customers understand the association between service applications and used resources.
s
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Topology Operation
 Beautify topology diagrams and perfect topology information.
 You can zoom in or zoom out a topology diagram, move an icon,

adjust the layout, and add or delete a virtual node in the
n
/e
topology diagram to make the topology diagram more eye-
pleasing and comply with user habits.
o m
 The following operations are supported: exporting and displaying
e i.c
a topology, retaining the previous status when a topology
aw
u
diagram is refreshed, deleting tips, switching an icon, replacing
h
an icon, and adding an application virtual node.
g .
ni n
ar
l e
: //
t tp
: h
e s
r c
o u
e s
R
in g
r n
e a
e L
o r
M
Network Topology
 View topologies based on the network connection dimension.
 A network topology refers to a physical layout in which devices are

interconnected using transmission media, that is, the way in which devices such
n
as computers in the network are connected. A topology diagram provides the
network configurations of network servers and workstations and connections
/e
between them.
o m
i.c
 The network topology of the operation and maintenance center can show
network connections between physical devices based on network layers
w e
ua
defined by the customer. The layers are the access layer, aggregation layer, and
h
core layer by default. The customer can also define connections between
devices.
g.
n i n
ar Page 23
l e
: //
ttp
 Network topologies help customers understand network connections of devices and support
fault locating.
: h
s
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Contents

n
o m
i.c

w e
hua
g.
n i n
ar Page 24
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Flexible Model Definition
 Customers are provided with a convenient and flexible object management model definition capability
and object models can be defined based on different O&M management requirements of customers.
For the model system that is actually running, an offline model editing capability by using a model
editing tool are also provided. Specific functions are as follows:
n
 Adding, modifying, and deleting a resource object.
/e
 Adding and modifying a resource attribute.
Defining the input mode of resource type attributes flexibly. At present, manual input, drop-down box selection,
m

o
time, and association are supported.
i.c
 Defining relationship between resource objects, including inclusion, classification, association, and dependency.
e
 Flexibly defining a resource object table index, including primary key index and multi-attribute unified index.
w
 Editing resource objects, object attributes, and relationship between objects offline.
ua
 Controlling whether the resource type is displayed in the left tree.
h
 Controlling whether different dimensions of the left tree are displayed.
g.
n i n
ar Page 25
l e
: //
ttp
 Object management models are edited based on service requirements.

: h
Adaptability of service models is improved to facilitate change and expansion of customers'
s
ce
O&M objects.
r
ou
es
When model definitions are edited, the inherent attributes on which services depend
cannot be modified.
R
n g
ni
Customized development is required for new relationship types.
r
Existing models can only be edited offline.
a
L e
e
or
M
Resource Maintenance
 Customers are provided with resource object maintenance management capabilities. Objects can be
maintained and managed through multiple modes such as manual entry, automatic discovery, and
import through Excel. Specific functions are as follows:
 Resource location definition: Support physical location management and logical location management; define
physical locations. The default location levels are province, data center, equipment room, and region.
n
/e
 Resource input: Support manual entry of objects and batch resource import, object review, manual combination,
service group definition, object type editing, and garbage resource deletion.
 Association of resource information with asset objects: Support customer entry and batch association of resources
o m
i.c
with customers.
Association of IP addresses with resources: Support entry of IP address segments and support associating devices
e

with IP addresses.
w
ua
 Data synchronization: Supporting manually accessing resources and querying third-party synchronization tasks.
 For resources that are automatically discovered, users are provided with the capability of reviewing
untrusted object devices.
. h
i n g
r n
l e a Page 26
: //
ttp
 Resource information is entered and maintenance management is performed for resource
information.
: h
s
ce
 Customers are provided with rich resource entry and management modes to facilitate O&M
management of users.
r

s ou
Restriction of use: During resource import and entry, at present, domain-based rights check
Re
(at present, administrators can enter resources beyond the authorized scope) is not
supported.
n g
r ni
e a
e L
or
M
Multi-dimension Object View
 Based on actual O&M requirements, O&M personnel can quickly locate and view resource
objects and perform related maintenance operations. Specific functions are as follows:
 Viewing objects based on the object category dimension.
 Viewing objects based on the physical location dimension (province, data center, equipment
room, and region).
n
 Viewing objects based on the virtual logic (resource zone and cluster) dimension.
/e
 Viewing objects based on the source system dimension.
o m
i.c
 Viewing objects based on the customer dimension.
 Viewing objects based on the service group dimension.
w e
ua
 Viewing objects based on the VDC dimension.
Viewing objects based on user-defined conditions.
h

g.
n i n
ar Page 27
l e
: //
ttp
 Provide views for checking objects from different angles of view.

: h
Facilitate O&M personnel to maintain objects in a routine way.
s
r ce
s ou
Re
n g
r ni
e a
e L
or
M
User-defined Object View
 O&M personnel can define required views based on requirements.
Specific functions are as follows:
 Defining a screening template in the object view and saving the template in
the view tree.
n
Selecting a service group and whether tenants are in the service group and
/e
m

are presented in the tenant view tree.

o
 Defining object resources offline and importing object resources into the
ei.c
w
system.
 Defining the display sequence of nodes in the object view offline.
hua
 Defining columns for displaying resource views.
g.
n i n
ar Page 28
l e
: //
ttp
 Edit screening conditions and define content of the object view to be checked.

: h
Quickly help O&M personnel meet requirements of different object maintenance scenarios.
s
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Comprehensive Presentation of Resource
Associated Information
 While viewing information about specific objects, O&M personnel can view multi-
dimension information such as alarm, topology, time, and performance related to the
objects and perform convenient O&M operations.
 In addition, the system supports viewing information about other associated objects from
n
/e
a single entry to facilitate maintenance personnel to quickly view more comprehensive
associated information while handling problems. Specific functions are as follows:
 Viewing the details, alarms, events, topologies, and performance information about objects when
o m
i.c
specific object data is selected.
 Determining whether to display the details, alarms, events, topologies, and performance
information when configuring different object types offline.
w e
ua
 Viewing information related to objects associated with the current object in object topology
information.
. h
i n g
r n
l e a Page 29
: //
ttp
 The system provides the capability of viewing multi-point information and the operation
capability.
: h
s
ce
 The system provides presentation of information about specific objects from multiple angles
r
to facilitate customers to maintain resource objects.
s ou
Re
n g
r ni
e a
e L
or
M
Resource Statistical Report
 The system provides the resource statistical report function and can collect statistics of
resources and export reports (in the PDF, RFT, CSV, Excel, Excel 2007, or Word format).
 Support statistical reports about resource quantities and ratios based on resource type. (Resource
types include server, firewall, switch, router, and array storage.)
Support reports about resource type distribution for the specified customer.
n
/e

 Support reports about details of VM resources allocated to customers, including user name, asset
number, private network IP address, elastic public network IP address, host IP address, VM
o m
i.c
specifications, and virtual management system.
w e
h ua
g.
n i n
ar Page 30
l e
: //
ttp
 The system provides the resource statistical report function and can collect statistics of
resources and export reports.
: h
s
ce
 The provided resource statistical report function helps users master the resource statistical
r
information about the whole data center.
s ou
Re
n g
r ni
e a
e L
or
M
Contents

n
o m
i.c

w e
hua
g.
n i n
ar Page 31
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Predefined Report
 Predefined templates include:
 Performance reports, including reports about top N server performance, top N network device
performance, top N network device port performance, top N array storage performance, CPU
utilization, top N CPU utilization, top N interface traffic, DC collation performance, and LUN IOPS
n
trend.
 Resource reports, including reports about device type statistics, customer assets, and asset details.
/e
 Capacity reports, including reports about VLAN resource pool capacity, public network IP address
o m
i.c
pool capacity, storage resource capacity, and computing resource capacity.
e
 Alarm reports, including reports about device alarm severity statistics, top N device alarm severity
statistics, device historical alarm severity statistics, top N device historical alarm severity statistics,
w
ua
alarm distribution diagram, alarm trend, and alarm trend comparison.
h
 Trouble ticket reports: including reports about trouble ticket overview, troubleshooting, fault
source proportion, and complaint proportion based on category.
g.
n i n
ar Page 32
l e
: //
ttp
 The system issues predefined report templates it carries.

: h
Predefined report templates help users quickly obtain common reports.
s
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Immediate Report
 Reports are provided and exported based on real-time data of
system operation.
 Output of objects and indicators of immediate reports is

n
/e
supported, including:
 CPU and memory reports of servers, VMs, and network devices
o m
 Interface traffic reports of network devices
ei.c
LUN IOPS reports of storage devices
w
ua

h
 Alarm statistical report
Trouble ticket report

g.
n

n i
ar Page 33
l e
: //
ttp
 Immediate reports help customers master real-time operation of devices in a data center.
: h
s
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Contents

n
o m
i.c

w e
hua
g.
n i n
ar Page 34
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Health Assessment of Cloud Data Centers
 Health assessment supports assessment at the following levels:
 Health assessment at the data center level
 Health assessment at the AZ level

n
 Health assessment at the cluster level
/e
 Health assessment at the DataStore level
o m
 Health assessment at the computing device level
ei.c
Health assessment at the storage device level
w
ua

 Health assessment at the network device level

. h
 Health assessment at the VM level
i n g
r n
l e a Page 35
: //
ttp
 Health assessment helps customers learn the health of data centers and helps customers
h
with proactive O&M based on health assessment suggestions.
:
s
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Health Diagram of Resource Pools
 Indicate the usage of virtual resource
Restriction of use:
n

 The feature is used only when supporting eSight and /e

FusionSphere are interconnected.
o m
ei.c
w
hua
g.
n i n
ar Page 36
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Risk Assessment of Cloud Data Centers
 Health assessment supports assessment at the following levels:
 Risk assessment at the data center level
 Risk assessment at the AZ level

n
 Risk assessment at the cluster level
/e
 Risk assessment at the DataStore level
o m
 Risk assessment at the computing device level
ei.c
Risk assessment at the storage device level
w
ua

 Risk assessment at the network device level

. h
 Risk assessment at the VM level
i n g
r n
l e a Page 37
: //
ttp
 Assess risks at various levels for computing resources, storage resources, network resources,
h
and virtual resources of cloud data centers based on the dimensions of remaining capacity,
:
s
resource consumption forecast (remaining time), and pressure, identify maximum risks, and
provide improvement suggestions.
r ce

s ou
Intuitively assess and forecast risks at various levels and help customers with proactive O&M
e
based on risk assessment suggestions.
R
n g
r ni
e a
e L
or
M
Efficiency Assessment of Cloud Data
Centers
 Efficiency assessment supports assessment at the following levels:
 Efficiency assessment at the data center level
 Efficiency assessment at the AZ level
n
/e
 Efficiency assessment at the cluster level
m
 Efficiency assessment at the DataStore level
o
i.c
 Efficiency assessment at the computing device level
 Efficiency assessment at the storage device level
w e
ua
 Efficiency assessment at the network device level
 Efficiency assessment at the VM level

. h
i n g
r n
l e a Page 38
: //
ttp
 Assess data centers at various levels for computing resources, storage resources, network
h
resources, and virtual resources of cloud data centers based on the dimensions of density,
:
s
inefficient operating status and low load distribution, and port utilization, identify inefficient
r ce
device resources, and provide improvement suggestions.
s ou
Intuitively assess and forecast efficiency at various levels and help customers with proactive
e
O&M based on efficiency assessment suggestions.
R
n g
r ni
e a
e L
or
M
Contents

n
o m
i.c

w e
hua
g.
n i n
ar Page 39
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Capacity Analysis
 Supported analysis views include:
 CPU, memory, and disk capacity at the data center level
 CPU, memory, and disk capacity trend at the resource pool level
 CPU, memory, disk, EIP, and VLAN capacity at the AZ level
n
/e
 CPU, memory, disk, EIP, and VLAN capacity at the cluster level
CPU, memory, disk, EIP, and VLAN capacity at the equipment room level
m

 CPU, memory, disk, and remaining network capacity at the data center level and risk assessment
o
e i.c
w
h ua
g.
n i n
ar Page 40
l e
: //
ttp
 Provide capacity snapshot views at various levels and trend analysis based on physical
location and virtual logic.
: h
s
ce
 Help customers learn capacity information and trends and implement proactive O&M.
r
s ou
Re
n g
r ni
e a
e L
or
M
Capacity Warning
 Based on preset capacity thresholds, capacity alarm notifications
are generated.
 Help customers discover and eliminate capacity risks in advance.

n
 Support customers in flexibly defining capacity thresholds (CPU
/e
crossing threshold, memory crossing threshold, and disk crossing
o m
threshold).
ei.c
Support generation of alarms and remind the administrator to
w
ua

h
perform expansion.
g.
n i n
ar Page 41
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Capacity Statistical Report
 Based on statistical capacity information, immediate capacity reports or
periodic capacity reports are provided and pushed to the specified customer.
 Supported capacity reports include:

 VLAN resource pool capacity reports (allocated amount and used amount)
n
 Public network IP resource pool capacity reports (total amount, allocated amount,
/e
and remaining amount)
o m
i.c
 Cluster storage resource pool capacity reports (total amount, allocated amount, and
remaining amount)
w e
ua
 Cluster computing resource pool capacity reports (total amount, allocated amount,
and remaining amount)
. h
i n g
r n
l e a Page 42
: //
ttp
 Capacity statistical reports help customers obtain statistical information about capacity of
resource pools.
: h
s
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Quiz
1. What are the focuses when viewing performance alarms?
2. What types of reports can be generated by the report management

function?
n
/e
o m
ei.c
w
hua
g.
n i n
ar Page 43
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Summary
 Management of alarms, performance, and resources
 Capacity management and health analysis of a resource pool
n
/e
o m
ei.c
w
hua
g.
n i n
ar Page 44
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Thank You n
www.huawei.com /e
o m
ei.c
w
hua
g.
ni n
ar
l e
: //
http
s :
r ce
sou
Re
n g
rni
e a
e L
or
M
Huawei Data Center
Computing Resource
Planning and Design
n
/e
o m
www.huawei.com
e i.c
w
hua
g.
ni n
ar
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Foreword
 In a data center, computing resources are indispensable
resources. Therefore, the planning and design of computing
resources become especially important. This chapter introduces
n
/e
computing resource planning and design by using cases.
o m
ei.c
w
hua
g.
n i n
ar Page 2
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Objectives
 Understand the overall structure of a data center.
 Master analysis on computing processing capabilities.

n
 Master design of computing scenarios.
/e
 Understand reliability design and security design.
o m
ei.c
w
hua
g.
n i n
ar Page 3
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Contents
1. Analysis on Computing Processing Capabilities
2. Design of Computing Scenarios
3. Reliability design
n
4. Security Design /e
o m
ei.c
w
hua
g.
n i n
ar Page 4
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
CPU-tpmC
 Calculation input: service transactions per minute (TASK), complexity ratio (S), CPU
utilization (C), service development redundancy (F), peak transaction time (T).
 Calculation formula: tpmC = TASK x S x F/(T x C)
Case: Assume a service system needs to calculate 2000 times of service access per second.
n

/e
In each minute, the service system calculates 120,000 times of service access. The peak
transaction time is one minute. Estimate the value of TPC-C of the database server within
five years. For example, the value of the empirical coefficient for retrieval and query is 7.5.
o m
 TPM = TASK x S x F/(T x C) = (2000 x 60) x 7.5 x [(1 + 0.3)5]/(1 x (1 - 0.5)) = 6683274 TPM
That is, a server of which the TPM value is not smaller than 6,683,274 is required.
ei.c
w

h ua
g.
n i n
ar Page 5
l e
: //
ttp
 TASK: service transactions per minute.

: h
S: ratio of complexity for querying service transactions to the standard TPC-C test reference
s
ce
environment. Because the actual complexity of querying service transactions differs greatly,
r
transactions in TPC-C standard test, a reasonable value must be set. Take common service
ou
transactions as an example. In one transaction, lots of database tables need to be opened
s
Re
and related data is taken out for operation. Relative to complexity of TPC-C standard
transactions, the complexity is higher and the general value ranges from 1 to 30 (a larger
n g
value indicates a more complex system).
r ni
C: processing margin of the host CPU. Practical application experience shows the CPU
a

e
utilization of a host server should not be higher than 75%. Based on service requirements, C
L
e
is generally set to 50%, that is, the processing margin of CPU is 1 - 0.5 = 0.5.
or  F: F is reserved for service volume development redundancy of the system in the coming n
M years. For applications that will be added successively in the future, 30% of processing margin
must be reserved. That is, for service development margin in the coming three years, the
calculation formula is as follows: (1 + 0.3)3 = 2.197.
 T: peak transaction time.

Memory
 For memory configuration, according to our past experience, memory is
consumed in the following parts:
 Memory consumed by normal operation of the host system (mainly
consumption of the OS)
n
Overhead required for database operation
/e
m

Normal overheads required for operation of database SGA

o
i.c

e
 Memory consumed by online transaction processing
w
ua
 Calculation formula: Memory = OS + Database management system +
h
Database SGA operation + Number of connections x 3 MB
g.
n i n
ar Page 6
l e
: //
ttp
 Case:
: h
We can see from above that the memory required by the server system consists of
s

ce
four parts. According to our previous experience, each connection occupies 2 MB to 3
r
MB memory. Generally, the OS occupies 500 MB memory, the database management
ou
system occupies 256 MB memory, the memory usage is not higher than 70%, the SGA
s
Re
of the retrieval system database requires 50 GB when it is running, and the number of
connections is 2000. The calculation formula is as follows:
n g
ni
 Memory = (512 MB/0.7 + 256 MB/0.7 + 2000 * 3 MB)/1024 + 50 GB = 59.45 GB
a rBased on the preceding formula, at least 64 GB memory is required.
L e
e
or
M
Virtualization Performance Calculation
 Virtualization loss: While improving the server usage effectiveness, the virtualization technology brings
performance loss. Based on the UVP and solution test results, it is recommended that the
virtualization loss be set to 10%.
 Logical core: The UVP schedules CPUs based on the unit of logical core. The logical core simulates a
n
physical core as two logical cores through the hyper-threading (HT) technology.
/e
 Calculation method for logical cores of computing nodes: Assume a server is configured with two 4-
m
core E5620 CPUs. There are eight (2 x 4) physical cores in total and 16 (8 x 2) HT logical cores. The
calculation methods for other servers are also the same.
o
i.c
 Available logical core: Some logical cores of computing nodes are exclusively occupied by Domain-0 of
w e
the UVP, and other logical cores are shared by all user VMs on the nodes. The shared logical cores are
available logical cores. In the current product version (FusionCompute R5C00), the UVP occupies two
ua
logical cores exclusively.
. h
i n g
r n
l e a Page 7
: //
ttp
 vCPU: virtual CPU. CPU virtualization means to provide each VM on the server with one or
more vCPUs.
: h
s
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Scenario in Which Logical Cores Are
Reused
 When logical cores are reused, the following conclusion is drawn from tests:
 For reuse of logical cores, when the service pressure on VMs is low, VMs invoke
internal vCPU resources within specifications and do not occupy all computing
resources within specifications. VMs with high service pressure will invoke resources
n
/e
of more physical cores to meet services. When idle resources of other services are
sufficient, VMs with high service pressure can exclusively occupy physical cores within
their specifications.
o m
 For reuse of logical cores, when lots of VMs meet high service pressure, the
performance of all VMs degrades sharply according to a certain law.
e i.c
w
ua
 The performance of VMs with higher specifications degrades more sharply;
. h
If more physical cores of the server are reused, performance degrades more sharply;
 If the traffic is heavier, the performance degrades more sharply.
i n g
r n
l e a Page 8
: //
ttp
 Reuse: If the total number of vCPUs of VMs running on computing nodes does not exceed the
h
number of available logical cores on the nodes, logical cores are not reused. On the contrary,
:
s
if the total number of vCPUs of VMs running on computing nodes exceeds the number of
r ce
available logical cores on the nodes, logical cores are reused.
s ou
Re
n g
r ni
e a
e L
or
M
Recommendation for Reuse of Logical
Cores
 Calculation formula: Total number of vCPUs of all VMs on computing
nodes/Number of available logical cores x 100%
 Based on actual test results, when resources such as memory and IO are met,
n
physical cores can be reused and more VMs are deployed on one CNA node to
increase the general utilization. The following table lists different vCPU reuse
/e
rates and corresponding applications scenarios:
o m
Application Scenario Recommended Reuse Rate
e i.c
w
vCPU utilization of VMs > 30% Not higher than 100% (that is, vCPUs are
ua
not reused)
20% < vCPU utilization of VMs < 30% 100% to 200%
. h
vCPU utilization of VMs < 20% 200% to 300%
i n g
r n
l e a Page 9
: //
ttp
 The reuse rate refers to the rate of all vCPUs of VMs on a server to available logical cores. It is
related to reuse of logical cores on the server.
: h
s
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Scenario in Which Logical Cores Are not
Reused
 When logical cores are not reused, the following conclusion is
drawn from tests:
 Multiple VMs are deployed on the CNA and physical cores are not
n
/e
reused. When other VMs on the CNA are pressurized to 100%,
the impact on the specified VM is lower than 10%. In this
o m
i.c
scenario, on one CNA node, the change of CPU pressure of other
VMs does not affect the specified VM.
w e
The computing capability of VMs increases in a linear way asathe

h u
number of VCPUs increases.
g .
n in
ar
e
/: /l
t tp
: h
e s
r c
o u
e s
R
in g
r n
e a
e L
o r
M
Calculating the Number of vCPUs
 How to calculate the number of vCPUs on a physical server? See
the following calculation formula:
 Number of vCPUs = (Number of physical CPUs x Number of physical
n
/e
cores x Number of logical threads for each core – Number of dom0
m
CPU cores) x Reuse rate
o
i.c
 Example:

w
Assume the server is configured with two 4-core E5620 CPUs, e
u a
Huawei UVP is deployed, and the recommended reuse rate is 150%.
Number of vCPUs = (2 x 4 x 2 - 2) x 150% = 21
. h
n g
r ni
l e a
: //
t tp
: h
e s
r c
o u
e s
R
g
r nin
e a
e L
or
M
Performance Baseline
 Because the capabilities of a vCPU are uncertain, a vCPU can provide a
maximum of one logical core (HT). As more VMs are deployed, the capabilities
of the vCPU degrade gradually. The capabilities of a vCPU also differ with the
dominant frequency of the CPU.
n
 Virtualization performance baseline:
/e
o m
i.c
Physical CPU Number of Value of Value of SPECfp_rate2006
e
Model Logical Cores SPECint_rate2006 for for Each Logical Core
Each Logical Core
w
ua
E5620 x 2 16 13.75 10.5
E5645 x 2 24 19.25
. h
14.63
i n g
r n
l e a Page 12
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Performance Conversion of P2V
 When a specific service is deployed on VMs or is migrated to VMs,
the service raises performance requirements. Generally, the
performance requirements are raised for physical servers. How
n
/e
to convert the performance requirements into requirements for
VMs?
o m
i.c
 The method for converting a CPU into a vCPU is as follows:
 Calculation input: CPU model of the physical server required by the
w e
service, CPU utilization, and model of CPUs on CNA nodes
hua
g.
n i n
ar Page 13
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Calculation Method
 Calculation method:
 Query the value of SPEC corresponding to the model of CPUs of the server required
by the service on the following website: http://www.spec.org/cpu2006/results. The
value is called PM_SPEC for short.
n
/e
 Query the virtualization performance baseline data table and obtain the vCPU
m
performance baseline of VMs based on the CPU configurations of CNA nodes.
o
i.c
 Number of vCPUs = PM_SPEC x CPU utilization x (1 + Redundancy factor)/vCPU
performance baseline
 The recommended value of the redundancy factor is 20%.

w e
hua
g.
n i n
ar Page 14
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Example
 Services on a PC are migrated to VMs. The PC is configured with
Intel Core i5-650 and the long-term CPU utilization is 30%.
 Query the website. The SPECint_rate2006 value of Intel Core i5-

n
/e
650 is 53. Based on the CPU configurations of CNA nodes of the
project, if the CPU is set to E5620, the queried vCPU performance
o m
i.c
baseline is 13.75.

w e
Number of vCPUs = 53 x 0.3 x (1 + 0.1)/13.75 = 1.27. Based on the
calculation result, two vCPUs need to be configured if the PCais
h u
migrated to VMs.
g .
n in
ar
e
/: /l
t tp
: h
e s
r c
o u
e s
R
in g
r n
e a
e L
o r
M
Virtualization - Homogeneous Server
 Conversion between vCPUs in calculation virtualization and physical CPUs.
Calculation virtualization involves homogeneous virtualization and
heterogeneous virtualization. Homogeneous virtualization involves servers on
Huawei platform and heterogeneous virtualization involves servers of other
n
/e
brands.
 Calculation input: CINT and CFP of physical CPUs, utilization of CPUs on the live
o m
i.c
network, and performance baseline value of Intel Xeon E5620

w e
Calculation formula: For homogeneous servers, if RH2285, T6000, and E6000 are used,
ua
use the following formula:
. h
i n g
r n
l e a Page 16
: //
ttp
 The performance reference described in this section is based on the preceding formula, in
h
which CINT and CFP are released in SpecCPU2006. For details, see
:
http://www.spec.org/cpu2006/results.
s
r ce
Physical CPU CINT: It refers to SpecCINT_rate2006. That is, performance indicator of the
ou

application that performs integer operations in the multi-processor computer system.
es
R
 Physical CPU CFP: It refers to SpecCFP_rate2006. That is, performance indicator of the
g
application that performs floating-point operations in the multi-processor computer system.
n
ni
 Service utilization: It generally refers to the CPU utilization of physical hosts in the migration
a r
scenario.

L e
Redundancy value: The recommended redundancy value ranges from 15% to 20%. The
e
or
redundancy value here considers the situation in which 13 vCPUs contend for resources. In
the earlier test, 10% of performance degradation occurs.
M
Virtualization - Heterogeneous Server
 For heterogeneous servers, the virtualization conversion
coefficient can be used for estimate.
n
/e
o m
ei.c
w
hua
g.
n i n
ar Page 17
l e
: //
ttp
 The performance reference described in this section is based on the preceding formula, in
h
which CINT and CFP are released in SpecCPU2006. For details, see
:
http://www.spec.org/cpu2006/results.
s
r ce
Physical CPU CINT: It refers to the SpecCINT_rate2006 value of the physical server for the
ou

application that bears services on the live network.
es
R
 Physical CPU CFP: It refers to the SpecCFP_rate2006 value of the physical server for the
g
application that bears services on the live network.
n
ni
 Heterogeneous server CPU CINT: SpecCINT_rate2006 value of the CPU of the physical server
a r
that bears virtual applications when the applications need to be switched to a virtualization
L e
platform.
e
or
 Heterogeneous server CPU CFT: SpecCFP_rate2006 value of the CPU of the physical server
that bears virtual applications when the applications need to be switched to a virtualization
M platform.
 Total number of cores of a heterogeneous server: Total number of cores of the physical
server that bears virtual applications (for example, if the physical server has two CPUs and
each CPU is configured with four cores, the total number of cores is: 2 x 4 x 4 = 16).
 Service utilization: It generally refers to the CPU utilization of physical hosts in the migration
scenario.
 Redundancy value: The recommended redundancy value is between 15% and 20%. The
redundancy value here considers the situation in which 13 vCPUs contend for resources. In
the earlier test, 10% of performance degradation occurs.
Example for Calculation of the Number of
Servers (1)
 The actual average CPU utilization of 107 Dell PowerEdge 2950 servers (configurations:
CPU 2 x E5420 @2.50 GHz, 4 Core(s), and 8.00 GB) is 20%. Table query shows that the
SPEC value is 118.
 After services are migrated to the RH5885 server (four 8-core CPUs, E7-4820, 2 GHz),
n
/e
query shows that the SPEC value is 775. Based on the SPEC value, the computing
capability of one vCPU of the server is as follows: SPEC cint2006 rates value of the server x
CPU utilization/(Number of CPUs x Number of cores x 2 - Number of logical cores
o m
i.c
consumed in virtualization) = 775 x 70%/(4 x 8 x 2 - 2) = 8.7.
 Number of required vCPUs = Roundup (118 x 20%/8.7) = 3. The required memory is 8 GB.
w e
h ua
g.
n i n
ar Page 18
l e
: //
ttp
 Theoretically, the factors that affect the number of servers are primarily the vCPU and
h
memory of VMs. The total number of servers is obtained based on the total number of vCPUs
:
s
and memory of VMs for service applications. At present, servers can use 16 GB and 32 GB
r ce
memory modules. Therefore, memory is no longer a bottleneck.
 Summary of VM resources
s ou

Re
Total number of VMs: 107
n g
Total number of vCPUs of VMs: 322
r ni
Total memory of VMs: 856 GB
e a
e L
or
M
Example for Calculation of the Number of
Servers (2)
 Based on the resource redundancy principle, the number of computing
resources required by the system is:
 Total number of vCPUs actually carried by the system: 322 x 120% = 387
n
 Total VM memory carried by the system: 856 GB x 120% = 1028 GB
 Based on the 30% configuration redundancy principle and selected servers (four 8-
/e
core CPUs), the number of servers and configuration are calculated as follows:
o m
i.c
 Number of servers calculated from the angle of vCPU = Total number of vCPUs/(Number of
e
CPUs x Number of CPU cores x 2 - 2) = Roundup (387 vCPUs/(4 x 8 x 2 - 2) = 7
w
 When 8 GB memory modules are used, the number of memory modules for each server is:
ua
(Total memory/Number of servers + 8 GB)/8 GB = (1028 GB/7 severs + 8 GB (consumed in
h
virtualization))/8 = 20
g.
n i n
ar Page 19
l e
: //
ttp
 To ensure reliability of VMs on the cloud platform, when a fault occurs on a server, make
h
VMs on the faulty server be able to run on other servers through the HA function provided by
:
s
the cloud platform and try to reduce the service interruption time. Therefore, when the
r ce
system is deployed, 20% (which can be changed depending on actual requirements)
ou
redundancy of the server CPUs and memory resources, as the computing resources of the
system, must be reserved.
es
R
n g
r ni
e a
e L
or
M
Contents
n
4. Security design /e
o m
ei.c
w
hua
g.
n i n
ar Page 20
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Selection of Physical Server (1)
Application Application CPU Memory Recommended
Common Service Scenario
Type Requirement Requirement Requirement Server Type
General basic
General
management system Two CPUs, Huawei
management
General type L L (Web, retrieval engine, X6000\E6000\RH228
application
DNS, DHCP, AD, FTP, and 5
system
n
FileServer)
/e
Basic tool application
system (for example, Two CPUs, Huawei
Tool application
Tool type L L printing control, report, X6000\E6000\RH228
m
system
OCR, streaming, and web 5
o
page capture)
i.c
Application Government portal
systems with Browse- website, weather query Four CPUs, Huawei
H H
e
heavy access intensive system, and Web RH5485
traffic middleware server
w
Two CPUs, Huawei
ua
Large IO, small Online processing
Application M H X6000\E6000\RH228
data amount database
systems with 5
large amount of
data Small IO, large
. hTwo CPUs, Huawei
g
M M Data warehouse analysis X6000\E6000\RH228
data amount
n
5
n i
ar Page 21
l e
: //
ttp
 As computer systems used by traditional data centers, physical servers can better show
h
performance of system software. The following introduces specific application scenarios of
:
physical servers in related scenarios.
s
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Selection of Physical Server (2)
Application CPU Memory Common Service Recommended Server
Application Type Need Requirement Requirement Scenario Type
Access/read- Two CPUs, Huawei

M H Video on demand
intensive X6000\E6000\RH2285
High-performance
computing cluster
n
/e
Application Database application
systems with Computing- Four CPUs, Huawei
H H server
m
large amount of intensive RH5485
o
dataz Digital urban
i.c
management
Image rendering
e
Streaming
Access/write- Four CPUs, Huawei
H H Database
w
intensive RH5485
Data warehouse
h ua
g.
n i n
ar Page 22
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Virtualization Adaptability
 For server requirements of the data center service in many sectors, first determine whether the server
supports a virtualization solution. If the server does not support a virtualization solution, adopt servers
that can meet actual requirements. It is recommended that others adopt virtualization in a unified way
and VMs of corresponding configurations are adopts to meet the requirements based on requirements
for collected or estimated computing resources.
n
/e
 In the following conditions, it is recommended that physical machines are directly used to meet
requirements of services for computing resources.
 Higher requirements are raised for operation performance of servers. When the VM with the maximum
o m
i.c
computing capability is configured on one physical server, the requirements for the computing capability of
e
service applications still cannot be met.
w
 Enterprise applications raise high requirements for the processing capabilities of video cards.
ua
 The encryption mode of the existing software licenses does not support the virtualization scenario.
h
 Service applications have special requirements for cards and cards do not support operation in the virtualized
environment.
g.
n i n
ar Page 23
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Analysis on Virtualization Adaptability
Network Storage
Application CPU Memory Storage Space Storage IO Virtualization
Application Type Bandwidth Bandwidth
Need Requirement Requirement Requirement Requirement Adaptability
Requirement Requirement
General
management General type L L L L L L Suitable
n
application system
Application systems
/e
Computing-
with large amount of H H M L M M Suitable
intensive
computing
Application systems
m
Browse-
with heavy access H H H M M L Suitable
intensive
o
traffic
i.c
Large IO, small Not recommended
M H M M H M
data amount generally
e
Small IO, large
Application systems data amount M M H H M H Suitable
w
with large amount of
ua
data Access/write- Not recommended
H H H H M H
intensive generally
h
Access/read-
M H H H M H Suitable
.
intensive
i n g
r n
l e a Page 24
: //
ttp
 For adaptability of application virtualization, the current mainstream application layering
h
(Web server, App server, and DB server) can be considered. Generally, virtualization suits
:
s
Web servers and App servers. DB servers need to be assessed based on requirements of
r ce
service applications for storage I/O. If DB servers of service applications require high I/O or
ou
the DB servers require the HA or cluster function, it is recommended that the DB servers of
s
the service applications are deployed on physical servers.
e

R
Virtualization supports Huawei cloud OS GalaX8800 on x86 servers of Huawei or mainstream
n g
vendors (IBM, HP, and Dell).
r ni
e a
e L
or
M
Contents
n
o m
ei.c
w
hua
g.
n i n
ar Page 25
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
OpenStack HA
 The reliability of OpenStack primarily refers to the reliability of
service functions provided by OpenStack, which covers the
following parts:
n
/e
 Rest API service reliability: Users are provided with uninterrupted
API services.
o m
i.c
 Database service reliability: Ensures user configuration data integrity
and service continuity.
w e
u a
Communication service reliability: Ensures uninterrupted interaction

between components.
. h
n g
r ni
l e a
: //
t tp
: h
e s
r c
o u
e s
R
g
r nin
e a
e L
or
M
Virtualization Reliability
 Virtualization is a core mechanism of data center. After a data
center is virtualized by using FusionSphere, the data center
supports the hot migration and HA functions of VMs and ensures
n
/e
continuity of service systems. The system supports the snapshot
function for VMs. It provides the VM and volume snapshot
o m
i.c
functions so that the system can be restored after the system
encounters a failure.
w e
hua
g.
n i n
ar Page 27
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Management Reliability
 FusionSphere management modules adopt 1+1 backup or load balancing during running. This ensures
reliability of system operation.
 FusionSphere supports monitoring for physical servers, software, and resources. The system can
collect core indicators of servers, for example, CPU utilization, basic network traffic, and memory data
n
and report alarms when the system detects faults such as abnormal process, abnormal management
/e
and storage link, abnormal node, and system resource overload. The alarms are generally cleared by
the system or handled by the administrator. In this case, the system maintains perfect fault detection
and fault recovery capabilities.
o m
i.c
 All physical servers of FusionSphere introduce the "black box" technology in the telecommunications
field. When the system becomes abnormal, kernel logs, system snapshots, kernel diagnosis
w e
information, and dying gasp are automatically stored and are saved in non-volatile devices (computing
ua
nodes) or are automatically transferred to the network server (for example, log server). In this case,
h
after the system fails, logs in the black box are exported and analyzed, the fault is quickly located, and
normal system operation is restored.
g.
n i n
ar Page 28
l e
: //
ttp
 Services in OpenStack adopt redundancy deployment, which involves the active-active mode
and the active-standby mode.
: h
s
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Server Reliability
 The cloud computing server supports real-time monitoring for temperature of
key thermal components such as CPU, memory, fans, power supply, and disks
and generates an alarm when a device fault occurs. Together with intelligent
fan speed regulation and monitoring, the server ensures reliable running of the
n
/e
server hardware system.
 The cloud computing server generally supports the hot swap and RAID
o m
i.c
functions of disks and provides online fault detection and warning for disks. It
supports 1+1 redundancy and hot swap of power supply. These measures
ensure the reliability of the server.
w e
hua
g.
n i n
ar Page 29
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Contents
n
o m
ei.c
w
hua
g.
n i n
ar Page 30
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Virtualization Security Threat
 In the enterprise data center, production data is deployed on the virtualization
platform. The current threats to security of the virtualization platform include:
 VMs attack Hypervisor.
n
 VMs attack and sniff each other.
 Threats caused by vulnerabilities of Hypervisor: The threats may cause VMs unable to
/e
provide services normally and destroy data security (confidentiality, integrity, and
o m
i.c
availability).
and losses to availability of virtualized networks.

w e
Viruses and worms: Viruses and worms bring losses to data integrity and availability
h ua
System configuration defects: Security defects exist in the system, which make
attacks, abuse, and misuse possible.
g.
n i n
ar Page 31
l e
: //
ttp
 While using benefits brought by the virtualization technology, enterprise users also face
h
security risks. The first point is whether the virtual layer can really securely isolate VMs from
:
s
hosts and isolate VMs from VMs. This is a fundamental point to ensure security of VMs. In
r ce
addition, malicious attacks between VMs inside cloud should be prevented. The network
ou
security protection devices in the traditional sense can no longer meet requirements for
s
protection of the virtualization layer.
e
R
n g
r ni
e a
e L
or
M
Host Security Design - Virus Protection
Design
 Deployment of host virus protection system
Internet Partner Branch
External connection zone
Internet Remote
n
Access zone LLB LLB
access zone
/e
FW FW FW FW LB
DNS FTP Web Server Farm
m
DMZ AV agent DMZ AV agent
o
i.c
Network
core layer
e
Production CSS
data center Core
w
ua
iStack iStack iStack
h
iStack iStack
.
FW
Web APP
g
iStack
n
Web APP
i
Server AV Server Server Server
AV agent Server AV agent DB
AV agent
DB DB
Operation AV agent
Development
n
management OA zone General service High-security service
zone and test zone production zone IP SAN存储区
production zone

ar Page 32
l e
: //
ttp
 As shown by the yellow icon in the figure, install an agent for hosts that requires virus
h
protection and manage the virus protection policy in a unified way through the AV Server
:
s
deployed in the operation management zone. Based on virus protection requirements of the
r ce
data center, provide servers based on Windows, Linux, and UNIX platforms with all-around
ou
virus protection capabilities to effectively defense important service servers, ensure
s
information security of the LAN, and make computers free from virus attacks.
e

R
Main product functions are as follows:
n g
ni
 Centralized network management: Provide a single management console to simplify
a r
management for Windows servers and networks.
L e
 Remote management: Remotely install, upgrade, and uninstall programs; update the
e
or
virus code file and the scanning engine and correct program download; scan and kill
viruses; install and set programs; provide real-time virus warnings and notifications;
M 
record and report virus events; and scan viruses in real time.
Update the virus code: The virus scanner can provide full effectiveness of virus
protection only after the latest anti-virus component is installed. Set automatic
download of the latest virus code and scanner engine and allocate the latest virus code
and scanner engine to the specified server. The new virus code adopts an intelligent
incremental update mode, that is, only the new virus code of the previous version
needs to be downloaded on one server. This efficient virus code update method saves
the download time and ensures smooth network bandwidth.
 Virus activity record report: A complete virus activity record report can trace and
manage lots of virus protection work, including discovery of virus infected file, update
of the virus code and programs, virus warnings, sending of suspiciously infected files
for examination, recording of the scanning time, and modification to write protection
for important directories.
n
/e
o m
e i.c
w
h ua
g.
n i n
ar
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Functional Design (1)
 VLAN isolation
n
/e
o m
ei.c
w
hua
g.
n i n
ar Page 34
l e
: //
ttp
 Based on the virtualization technology, Huawei SingleCloud platform is deployed in the
h
enterprise data center for security threats the virtualization layer faces. This is to provide the
:
s
virtualization function through Hypervisor and provides the cloud platform management
r ce
function through Galax8000 platform. The following describes security features provided by
ou
the SingleCloud platform, which are used in functional design of the virtualization security
s
infrastructure in the enterprise data center solution.
e

R
The virtual switching function is implemented through a virtual bridge. The VLAN tagging
n g
function supported by virtual bridge implements VLAN isolation and ensures security
r ni
isolation between VMs, so as to ensure communication data security between VMs running
a
on a physical server.
e

e L
A bridge supports the VLAN tagging function. VM instances of one virtual security group
or
distributed on multiple physical machines can identify data frames through VLAN tagging.
M
Switches and routers on the network can decide to route and forward data frames based on
VLAN ID. The virtual network isolation function is provided.
Functional Design (2)
 Network isolation of security groups
 A VM security group is the collection of VMs, namely collection of network

security rules about the VMs. VMs in one VM security group may be distributed
n
on physical machines in different physical locations. Therefore, a VM security
group is used to allocate a mutually isolated logical LAN in a physical network to
/e
improve network security.
o m
i.c
 This function allows end users to control interworking between their own VMs
and their other VMs or between their own VMs and VMs of other personnel.
w e
ua
The interworking between VMs is implemented by configuring interworking
h
rules between security groups.
g.
n i n
ar Page 35
l e
: //
ttp
 One user can create multiple security groups but one security group belongs to only one user.
h
When a user creates a VM, the user can determine the security group of the VM.
:
s
ce
 By default, VMs in one security group support interworking. By default, VMs in different
r
security groups are completely isolated.

s ou
Security group rules belong to one-way white list rules. A user can allow VMs in one of the
Re
user's security groups to receive requests from VMs in other security groups or requests from
g
an IP address segment. Request types are also configurable, for example, TCP and ICMP
n
ni
requests.

a r
Security group rules take effect after VMs are started and are migrated among computing
L e
servers after VMs are migrated. Users only need to set rules without worry about where VMs
e
or
are running.
M
Quiz
1. What factors need to be considered for virtualization performance
calculation?
2. How do you calculate the number of vCPUs?
n
/e
3. What factors are involved in reliability and security design?
o m
ei.c
w
hua
g.
n i n
ar Page 36
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Summary
 The slides describe planning and design of a DC's computing
resources in the following sequence: analysis on computing
processing capabilities, design of computing scenarios, reliability
n
/e
design, and security design.
o m
ei.c
w
hua
g.
n i n
ar Page 37
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Thank You n
www.huawei.com /e
o m
ei.c
w
hua
g.
ni n
ar
l e
: //
http
s :
r ce
sou
Re
n g
rni
e a
e L
or
M
Huawei Data Center Storage
Resource Planning
n
/e
o m
www.huawei.com
e i.c
w
hua
g.
ni n
ar
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Foreword
 A data center consists of the professional storage layer, storage
devices, and software. Therefore, a proper plan of the storage
resources is particularly important. In this course, certain cases
n
/e
will be presented to introduce the planning of storage resources.
o m
ei.c
w
hua
g.
n i n
ar Page 2
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Objectives
 On completion of this course, you will be able to:
 Analyze storage processing capability.
 Calculate the required Huawei storage product resources.

n
 Design storage scenarios.
/e
o m
ei.c
w
hua
g.
n i n
ar Page 3
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Contents
1. Storage Processing Capability Analysis
2. Calculation of Required Huawei Storage Product Resources
3. Storage Scenario Design

n
/e
o m
ei.c
w
hua
g.
n i n
ar Page 4
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Storage Processing Capability Analysis
 Generally, storage requirements are based on three dimensions:
 Storage performance
 Storage architecture
n
 Storage capacity
/e
 To meet the requirement for storage resources, the IOPS, storage
o m
capacity, and storage bandwidth of the storage devices are
ei.c
calculated.
w
hua
g.
n i n
ar Page 5
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Storage Performance
 The storage performance specification (specifically, the IOPS) is the basis for choosing a
storage product.
 The cache hit ratio, which changes with the actual service, is not considered as a selection
criteria.
n
/e
 The calculation involves the server tpmC, I/O empirical value, and redundancy for future
development.
o m
i.c
 The calculation formula is as follows:
e
 Storage performance (IOPS) = tpmC/T x S x F.
w
Assume that the server tpmC is 1 million for accommodating storage requirements for the
ua
next 3 years, then the calculation formula is as follows:
 Storage IOPS = 100 million/60 x 50 x [(1 + 20%)3] = 144 million

. h
i n g
r n
l e a Page 6
: //
ttp
 The tpmC is a server performance indicator.

: h
T indicates the service transaction volume per minute. If each transaction costs one second,
s
ce
then T = 60.
r
ou
 S indicates the empirical IOPS value. According to the industry experience, the tpmC
s
complexity is related to the transaction connotation, transaction process, hardware
e
R
configuration, and software efficiency. Generally, one tpmC complexity consists of about 20
g
to 30 I/O operations, such as read, write, and modification operations. The tpmC complexity
n
ni
is set to 30 I/O operations. The batch service processing of statistic query operations will be
r
supported later; therefore, the I/O value is high. According to industry experience, the I/O
a
e
value ranges from 80 to 100. Therefore, the overall I/O value is set to 50.
L
e
or
 F specifies the redundancy for future development. The growth of services and user data
affects the purchased storage devices. Therefore, during storage configuration, the
M development assessment for the next few years is required. Generally, the development rate
is 10% to 30% per year. The comprehensive value is set to 20%. That is, if the storage is
planned for the next five years, then F = (1 + 20%)3 = 2.49.
Storage Architecture
 The storage architecture is divided into IP SAN, Fibre Channel SAN, and NAS.
The storage plan is made mainly based on IP SAN and Fibre Channel SAN.
 Actual bandwidth = B x U x L/S
 The IP SAN bandwidth usage is 60%. The header loss rate is 20%. Therefore, the
n
gigabit bandwidth usage of the IP SAN is as follows:
/e
 1000 Mbit/s x 60% x (1 – 20%)/8 = 60 Mbit/s
o m
 The Fibre Channel SAN bandwidth usage is 80%. Therefore, the Fibre Channel
SAN 4 Gbit/s bandwidth usage is as follows: ei.c
w
ua
 4000 Mbit/s x 80%/8 = 400 Mbit/s
. h
i n g
r n
l e a Page 7
: //
ttp
 Architecture measurement standard. Application databases can be accessed through OLTP or
h
OLAP. The OLTP supports random read/write operations, and the OLAP supports sequential
:
s
read/write operations. In large scale applications, the application rate of the OLTP is nine
r ce
times that of the OLAP. In the Oracle database, the default size of a data block is 8 KB,
ou
whereas a data block of 64 KB is required for the OLAP.

es
B specifies the mainstream bandwidth value of the IP SAN or Fibre Channel SAN. Based on
R
features of the IP SAN and Fibre Channel SAN, the bandwidth of the mainstream network
n g
architecture IP SAN is 1000 Mbit/s, and the bandwidth of the Fibre Channel SAN is 4000
Mbit/s.
r ni

e a
U specifies the bandwidth usage. Neither the IP SAN nor Fibre Channel SAN can use 100%
e L
bandwidth during data transmission. Therefore, this parameter is the actual bandwidth usage
or
during data transmission. The value is 80% for the Fibre Channel SAN and 60% for the IP SAN.
M  L specifies the header loss rate. During data encapsulation, loss of the transmission efficiency
is caused. The value of this parameter is the bandwidth loss rate during data encapsulation.
The bandwidth loss rate is usually 1 (no loss) for the Fibre Channel SAN and 20% for the IP
SAN.
 S specifies the conversion value between the bit and byte. One byte equals eight bits. The
value of this parameter is 8.
Calculation of Required Bandwidth
 Inputs: Number of online concurrent users, and proportions of OLTP and OLAP.
 Formula: Required bandwidth = TASK x (X% x OLTP data block size + Y% x OLAP
data block size)
 Assume an enterprise with 22500 staff. If the number of online concurrent

n
users per second is 2250, each submitting 2 transaction requests per second,
/e
then
o m
 Required bandwidth = 2250 x 2 x (90% x 8 KB + 10% x 64 KB) = 59.8 Mbit/s.
ei.c
w
 According to the calculations based on storage performance and architecture,
ua
IP SAN is recommended for this case.
. h
i n g
r n
l e a Page 8
: //
ttp
 Architecture measurement standard. Application databases can be accessed through OLTP or
h
OLAP. The OLTP supports random read/write operations, and the OLAP supports sequential
:
s
read/write operations. In large scale applications, the application rate of the OLTP is nine
r ce
times that of the OLAP; therefore, X = 90, and Y = 10. In the Oracle database, the default size
ou
of a data block is 8 KB, whereas a data block of 64 KB is required for the OLAP.
es
R
n g
r ni
e a
e L
or
M
Calculation of Device Storage Capacity
 Inputs: Nominal capacity of a disk, RAID level and planning, hot spare disk
configuration, coffer disk capacity, and total number of disks.
 Formula: Required capacity = Total capacity of current applications/Storage
n
redundancy * Required capacity for future development
 If the total capacity of current applications of an enterprise is 10 TB, then

/e
Required capacity for future three years = 10 T/(1 - 0.35) x ((1 + 20%)3) = 26.59
o m
i.c
TB
w e
hua
g.
n i n
ar Page 9
l e
: //
ttp
 Total capacity of current applications indicates the total storage capacity of all service
systems in an enterprise.
: h
s
ce
 Storage redundancy = 1 - loss rate of the disk array caused by RAID and hot spare disk
configuration (usually 35%)
r

s ou
The development of an enterprise brings larger numbers of both the staff and customers,
Re
and consequently a higher required storage capacity. If the storage capacity increases by 20%
g
every year, then the required capacity for future three years is calculated as follows: (1 +
n
ni
20%)3 = 1.728.
a r
L e
e
or
M
Contents

n
/e
o m
ei.c
w
hua
g.
n i n
ar Page 10
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Device IOPS
 The cache hit ratio is not involved in the calculation since it changes with the actual
service.
 Inputs: IOPS of a disk, RAID level, write ratio, and total number of disks.
Calculation:
n

/e
 Calculate the valid IOPS ratio based on the RAID level and write ratio.
 Valid IOPS ratio of RAID5 = 1/(1 + 3 x write ratio)
o m
i.c
 Valid IOPS ratio of RAID6 = 1/(1 + 5 x write ratio)
 Valid IOPS ratio of RAID10 = 1/(1 + write ratio)
w e
ua
 Calculate the device IOPS.
h
 Device IOPS = IOPS of a disk x total number of disks x valid IOPS ratio
g.
n i n
ar Page 11
l e
: //
ttp
 One write I/O operation on data disk D involves the following four steps: 1. Read data from
h
data disk D. 2. Read data from parity disk P. 3. Write data to parity disk P. 4. Write data to
:
data disk D.
s
r ce
As you can see, a write I/O operation actually ends up with two write I/O operations and two
ou

read I/O operations, which means three more I/O operations. However, a read I/O operation
es
does not lead to extra I/O operations.
R

n g
Assume that the write I/O operations take up x% in all I/O operations, then the valid IOPS
ni
ratio = 1/(1 + 3x%), that is, 100/(100 + 3x).
a r
L e
e
or
M
IOPS Calculation Example
 Take the service model of an online video media server as an example
to calculate the storage IOPS. The I/O operations of the server are
mainly read operations, and write operations count only 5%.
n
Total
/e
m
IOPS of a Write
Number of RAID Level Calculation
o
Disk I/O ratio
Disks
RAID5
RAID6
140 x 12/(1 + 3 x 5%) = 1460
140 x 12/(1 + 5 x 5%) = 1344
e i.c
140 5% 12
w
RAID10
ua
140 x 12/(1 + 5%) = 1600
h
g.
n i n
ar Page 12
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Calculation of Device Storage Capacity (1)
 Configuration of hot spare disks for RAID 2.0+
Number of Each Type of Number of Hot Spare Disks with Equivalent Capacity
n
Hard Disks in the Disk
/e
Domain
High hot spare Low hot spare
(0, 12] 1 1
o m
i.c
(12, 24] 2 1
(24, 48] 3 2
(48, 72] 4 2
w e
h ua
g.
n i n
ar Page 13
l e
: //
ttp
 RAID 2.0+ uses distributed hot spare space to replace the traditional hot spare disks. The
h
storage system automatically reserves hot spare space according to the hot spare policy so
:
s
that user operations are not needed. When the system automatically detects an
r ce
uncorrectable medium error on a hard disk or the entire hard disk is faulty, automatic system
ou
reconstruction will move affected data blocks to the hot spare space of normal hard disks. In
s
this way, the system self heals quickly.
e

R
For every 96 extra disks after the total number of hard disks exceeds 168, one hot spare disk
n g
is added under high hot spare policy, and the number of hot spare disks under low hot spare
r ni
policy is half of that under high hot spare policy (the number with decimal rounds up to an
a
integer).
e

e L
This table only applies to a single disk domain. If multiple disk domains or multiple layers of a
or
single disk domain are involved, the hot spare space must be calculated separately for each
M
disk domain or layer.
 The total hot spare space of a disk domain is calculated using the hot spare space for each
type of disks in the domain. The total hot spare space of a storage system is the sum of hot
spare space of all disk domains.
Calculation of Device Storage Capacity (2)
 Configuration of hot spare disks for RAID 2.0+
(72, 120] 5 3
(120, 168] 6 3
n
/e
(168, 264] 7 4
(264, 360] 8 4
o m
i.c
(360, 456] 9 5
(456, 552] 10 5
w e
ua
(552, 648] 11 6
(648, 744] 12
.
6
h
i n g
r n
l e a Page 14
: //
ttp
 Examples
: h
Assume a disk domain contains 96 SAS disks. According to the table, the number of hot
s

ce
spare disks with equivalent capacity under high hot spare policy is 5.
r
ou
 If a storage system contains 192 SAS disks, and the disks are divided into two domains
s
with 96 disks each, then the number of hot spare disks is 10, which differs from the
e
R
number (6) in the table.
n g
r ni
e a
e L
or
M
Main Factors Affecting the Disk Utilization
Ratio (1)
 RAID capacity loss: Different RAID levels differ in the RAID algorithm, leading to
different utilization ratios of disk capacity. For example, the utilization ratio of
RAID10 is merely 50%, whereas the utilization ratio of RAID5 is (n-1)/n.
n
 Difference between the available capacity in the OS and disk nominal capacity:
The hard disk vendors use 1000 as the conversion ratio (1 KB = 1000 bytes, 1
/e
MB = 1000 KB, 1 GB = 1000 MB), whereas the OS uses 1024 as the conversion
o m
i.c
ratio (1 KB = 1024 bytes, 1 MB = 1024 KB, 1 GB = 1024 MB). Therefore, the
nominal capacity is larger than the available capacity in the OS. The larger the
disk capacity is, the larger the difference between the two values is. The
w e
available capacity in the OS is probably 93.1% of the nominal capacity.
hua
g.
n i n
ar Page 15
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Main Factors Affecting the Disk Utilization
Ratio (2)
 Writehole space: To improve the RAID data reliability, 256 MB is
reserved on each disk.
 Coffer space: When a disk is used as a coffer disk, some space is used to
store certain service data and configuration data. The coffer space is 5
n
GB for OceanStor 18000 and OceanStor V3. /e
o m
i.c
 Metadata: The metadata counts 5.2% of the OceanStor capacity.
w e
hua
g.
n i n
ar Page 16
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
RAID 2.0+ Disk Capacity Utilization Ratio
 Not considering the capacity consumption of hot spare, the RAID
2.0+ disk capacity utilization ratio is calculated as follows:
 RAID 2.0+ disk capacity utilization ratio = Traditional RAID disk
n
/e
capacity utilization ratio x (1 – 5.2%) x (1 – 1%)
 The available capacity of RAID 2.0+ is calculated as follows:

o m
 Available capacity = (Total capacity – Hot spare capacity – Number
ei.c
w
of coffer disks x Coffer capacity) x RAID 2.0+ Disk capacity utilization
ua
ratio
. h
i n g
r n
l e a Page 17
: //
ttp
If a system contains 48 SAS disks with nominal capacity of 600 GB and 4 coffer disks, and RAID5
(8D+1P) policy is adopted, then
: h
s
ce
 Available capacity of the OS = 600 GB x 0.931 = 558.6 GB
r
ou
 The integrated capacity of a disk is about 558 GB for a 512B disk and 549 GB (558 GB x 98.4%)
for a 520B disk.
es

R
Total raw capacity = Total number of disks x Capacity of an integrated disk is as follows:
n g
ni
512B disk: 48 x 558 GB = 26.156 TB
r
520B disk: 48 x 549 GB = 25.734 TB
a
L e
The capacity utilization ratio is 83.42%.
e
or
M
Calculation of Storage Device Bandwidth
 Inputs: total interface bandwidth and redundancy coefficient (15%
recommended)
 Formula: Storage device bandwidth = Total interface bandwidth x (1 –
n
Redundancy coefficient)
 Calculation example:
/e
 An IP SAN device with eight 1 GE iSCSI interfaces has a total bandwidth of 8 Gbit/s. If
o m
i.c
the redundancy coefficient is 15%, then
 Device bandwidth = 8 x (1 – 15%) = 6.8 Gbit/s
w e
ua
 A Fibre Channel SAN device with four 8 Gbit/s fibre channel interfaces has a total
h
bandwidth of 32 Gbit/s. If the redundancy coefficient is 15%, then Device bandwidth
= 32 x (1 - 15%) = 27.2 Gbit/s
g.
n i n
ar Page 18
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Contents

n
/e
o m
ei.c
w
hua
g.
n i n
ar Page 19
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Fibre Channel SAN
 The Fibre Channel SAN is mainly used in large-size database server systems
(such as Oracle, DB2, and Sybase) and database servers deployed in cluster.
 High performance, security, and stability are required for critical services.
 The service demand drives the performance requirement. The purchased

n
devices can meet the performance requirement at first. However, as the service
/e
requirement greatly increases, the underlying storage requirement increases
o m
i.c
quickly. The traditional storage system fails to meet the increasing
requirements.
w e
ua
 The tiered storage solution uses application servers, features a large overall
system capacity, and has high concurrent access requirements. Generally, 20%
. h
i n g
of service data is frequently accessed, while the other 80% service data is idle.
r n
l e a Page 20
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
IP SAN
 The IP SAN is mainly used in medium- and small-size database server systems
(such as SQLServer, MySQL, and PGSQL), mail server, DNS server, file server,
and WINS server.
n
 Non-critical services do not require high data read/write capabilities.
 Critical services do not require high data read/write capabilities.

/e
 Deployment of a large number of VMs: With the rapid development of
o m
computing virtualization, many non-core application systems and virtual
desktops are deployed on VMs, which results in higher requirements in storage ei.c
w
ua
capacity, and storage performance and scalability.
. h
i n g
r n
l e a Page 21
: //
ttp
 The NAS is recommended for unstructured data, such as the office document, text, figure,
h
Extensible Markup Language (XML), Hypertext Markup Language (HTML), report, picture, and
:
audio/video information.
s
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Recommended Storage for Different
Applications (1)
Storage Storage
Application Application Storage I/O Common Service Recommended
Space Bandwidth
System Feature Requirement Demand Scenario Storage
Demand Demand
IP SAN, Huawei
n
General Basic management
Common L L L S5500T (SATA
/e
management systems
disks)
Large
Computing-
Teachers’ scientific
research projects (high-
IP SAN, Huawei
o m
i.c
computing L M M S5500T (SATA
intensive performance computing
volume disks)
and graphics rendering)
Net schools and home-

w e
IP SAN, Huawei
ua
Many access Browse-
M M L school interconnection S5600T (SATA
requests intensive
projects disks)
Large data
Large I/O and
Online transaction
. h IP SAN, Huawei
g
small data M H M
volume processing databases S5500T (SAS disks)
n
amount
n i
ar Page 22
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Recommended Storage for Different
Applications (2)
Application Storage Storage
Application Storage I/O
System Space Bandwidth Common Service Scenario Recommended Storage
Requirement Demand
Feature Demand Demand
Large I/O and small Online transaction processing
M H M IP SAN and Huawei S5500T (SAS disks)
data amount databases
Small I/O and large IP SAN & Fibre Channel SAN, Huawei
H M H Analysis of data warehouse
n
data amount S5600T & S5800T (SSD/FC disks)
Large data
/e
Access-intensive Video uploading websites and online IP SAN & Fibre Channel SAN, Huawei
volume H M H
and write-intensive hard disks S5800T & S6800T (SAS/FC/SSD disks)
m
Video On Demand systems
Access-intensive IP SAN & Fibre Channel SAN, Huawei
H M H Multimedia teaching
o
and read-intensive S5800T & S6800T (SAS/FC/SSD disks)
Education blogs
i.c
Office documents, texts, pictures,
Unstructured
e
L L L XML, HTML, reports of various types, N8500 (SATA disks)
data
images, and audio/video information
Large-scale critical service systems
w
ua
Special
Other ¹ H H H such as large-scale retrieval systems, HP XP24000 & EMC VMAX
requirements
and BOSS & CRM & BASS
. h
i n g
r n
l e a Page 23
: //
ttp
 H: high; M: medium; L: low.
 Other ¹ indicates users’ special needs in three aspects:

: h
s

r ce
Inheritance: The user system needs to be upgraded. To reuse resources during upgrade,
ou
third-party high-end disk array deployment is used.
es
User-specified: The core application systems of most user service systems use the third-party
R
high-end disk array for storage. Therefore, when creating a service system, considering the
g
high requirements on technicians’ familiarity with the array and the array stability, this array
n
ni
can also be chosen.
a r
e
 Computing capability: According to the calculation results of IOPS, bandwidth, and capacity,
e L
select the disk array that meets the requirements.
or
M
Storage Virtualization
 The storage virtualization feature can virtualize storage devices of various brands and models into a unified storage
resource pool. After storage devices are virtualized into a storage resource pool, existing storage devices can be flexibly
used and storage devices can be added. After virtualization, the existing storage devices can be reused. Supported
operations include data integration, migration, mirroring, and remote replication.
n
/e
HOST
HOST Cloud Non-cloud
scenario scenario
o m
i.c
Switch Switch
Storage
plane
VIS6000
w e
h ua
.
IP SAN FC SAN IP SAN
Huawei
Storage
Heterogeneous storage
i n g
r n
l e a Page 24
: //
ttp
 Storage virtualization is applicable to the following scenarios:
: h
The openness of heterogeneous SAN or IP SAN storage system is required.
s

 Reusing storage products.

r ce
ou
 The effective data migration between heterogeneous storage systems is required.
es
Resource management of heterogeneous storage systems is required.
R

n g
Integration of servers and storage.
r ni
Performance improvement of storage pools.
e

a Data protection and restoration in a heterogeneous storage environment.
e L Disaster recovery (DR) of heterogeneous storage.
or

 For the preceding scenarios, Huawei VIS6000 storage virtualization product solutions are
M recommended.
Distributed Storage
Storage SCSI driver/iSCSI Volume
driver layer mgmt
Distributed linked clone Capacity

Backup mgmt
Storage Distributed snapshot
n
System
service layer Distributed thin provisioning
/e
Disaster monitoring
Distributed cache recovery

Storage
m
mgmt
o
Storage Cluster status control Data consistency management
Cluster faults
i.c
engine layer self-healing Unified
Data route management Parallel data reconstruction alarm
Hardware
E9000 computing and storage converged blade server
x86 CPU SAS/SATA PCI-E SSD card
w e
QoS
ua
mgmt
device layer
10 GE/InfiniBand
. h
i n g
r n
l e a Page 25
: //
ttp
 FusionStorage is a distributed storage software product developed by Huawei. It uses
h
innovative architecture and design and features high performance, reliability, and cost-
:
s
effectiveness. It highly integrates storage and computing and offers consistent and predicable
r ce
performance, scalability, flexibility, and self-healing. FusionStorage uses the distributed
ou
cluster control and hash routing technologies to provide distributed storage.

es
Storage interface layer: provides volumes for operating systems and databases over the Small
R
Computer System Interface (SCSI).
n g
ni
 Storage service layer: provides various advanced storage features, such as snapshot, linked
r
cloning, thin provisioning, distributed cache, and backup and disaster recovery (DR).
a

L e
Storage engine layer: provides basic storage functions, including management status control,
e
or
distributed data routing, strong-consistency replication, cluster self-healing, and parallel data
reconstruction.
M  Storage management layer: provides the operation and maintenance (O&M) functions, such
as software installation, automatic configuration, online upgrade, alarm reporting,
monitoring, and logging, and also provides a portal for user operations.
Principle of FusionStorage Distributed
Storage Resource Pool
Server 1 Server 2 Server N
VM VM VM VM VM VM
n
/e
...
o m
i.c
FusionStorage Distributed Storage Resource Pool
w e
ua
SSD HDD SSD HDD SSD HDD
. h
i n g
r n
l e a Page 26
: //
ttp
 Huawei distributed cloud data center solution uses the FusionStorage system. The
h
FusionStorage uses the new-generation distributed storage architecture and parallel
:
s
distributed grid storage technology. The distributed multi-node grid shares storage loads
r ce
concurrently. The fine-grained data distribution algorithm ensures that data is distributed
ou
evenly. The FusionStorage improves system reliability, availability, and access efficiency. The
s
FusionStorage capacity is easy to expand. Simply speaking, the FusionStorage can be
e
R
deployed on common servers to consolidate local disks on all servers into a virtual storage
g
resource pool. Volumes are fragmented and distributed to all hard disks in the resource pool,
n
ni
thereby achieving fine-grained, high-concurrency data storage and retrieval.
a r
L e
e
or
M
Quiz
1. From which dimensions are demands for storage capabilities posed?
2. From which dimensions are demands for storage resources posed?
n
/e
o m
ei.c
w
hua
g.
n i n
ar Page 27
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Summary
 This course introduces the three aspects involved in the storage
resource planning:
 Storage processing capability analysis
n
/e
 Calculation of required storage resources
 Storage scenario design

o m
ei.c
w
hua
g.
n i n
ar Page 28
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Thank You n
www.huawei.com /e
o m
ei.c
w
hua
g.
ni n
ar
l e
: //
http
s :
r ce
sou
Re
n g
rni
e a
e L
or
M
Huawei Distributed
Storage Solution
n
/e
o m
www.huawei.com
e i.c
w
hua
g.
ni n
r
l e a
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Foreword
 This document describes the basics of FusionStorage, including its
concept, application scenarios, compatibility, and logical
architecture. As a start of the whole course, this document aims
n
/e
to let trainees have a preliminary understanding of FusionStorage,
laying a solid foundation for subsequent learning.
o m
ei.c
w
hua
g.
n i n
ar Page 2
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Objectives
 On completion of this course, you will be able to:
 Describe FusionStorage and its application scenarios.
 Master the FusionStorage compatibility and license information.

n
 Master and describe the FusionStorage logical architecture.
/e
 Understand the FusionStorage DR solution.
o m
ei.c
w
hua
g.
n i n
ar Page 3
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Contents
1. Product Overview
2. Basic Concepts and Key Principles
3. Component Functions and Mutual Relationships

n
4. I/O Process Analysis /e
o m
i.c
5. Main Functions and Features
w e
hua
g.
n i n
ar Page 4
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Server SAN
 Concept
 A server SAN is a storage resource pool which consists of the storage
resources of multiple independent servers. It integrates both
n
/e
computing and storage resources.
 Features
o m
 Dedicated devices functioning as universal devices
ei.c
 Linear expansion of computing and storage
w
 Simple management and low TCO
hua
g.
n i n
ar Page 5
l e
: //
ttp
 Decoupled from vendor-specific hardware: Hardware and software are highly integrated in
h
traditional storage systems, whereas server SAN products are decoupled from hardware and
:
not bound to specific hardware.
s
r ce
Converging storage and computing: The server SAN is built on a universal x86 server,
ou

converging both computing and storage.
es
R
n g
r ni
e a
e L
or
M
Huawei Server SAN Product:
FusionStorage
 Distributed block storage software
 Employs distributed technologies to organize hard disk drives (HDDs) and solid-state drives (SSDs) of
x86 servers into a large-scale storage resource pool.
 Provides standard SCSI and iSCSI for upper-layer applications in non-virtualization environments and
n
virtual machines (VMs).
/e
Computing
 Open APIs
m
Storage Storage
o
PCIe Controller PCIe
Computing Computing
ei.c
w
ua
Controller Controller
. h
PCIe
PCIe
Storage Storage
Controller
Computing
i n g
r n
l e a Page 6
: //
ttp
 Supports typical application scenarios of the traditional block storage, such as SQL, Oracle
RAC, Web, and industry applications.
: h
s
ce
 Supports integration with mainstream cloud platforms, such as Huawei FusionSphere,
r
VMware, and open-source OpenStack. Storage resources can be allocated on demand.

s ou
FusionStorage is the only commercial server SAN product that supports PB-level data
processing.
Re
n g
r ni
e a
e L
or
M
Traditional SAN Architecture
Server Server
Controller Controller Controller Controller
VM VM VM VM Controller Controller Controller Controller
n
VM VM VM VM
/e
m
SSD HDD SSD HDD SSD HDD Disk enclosure
FC/IP Network FC/IP Network
o
i.c
......
Cache HDD Cache HDD
SSD HDD SSD HDD SSD HDD
w e
Disk enclosure
ua
SAN/NAS SAN/NAS
Only capacities of storage devices
h
Isolated storage resources
can be expanded.
g.
n i n
ar Page 7
l e
: //
ttp
 Traditional storage resources cannot be shared. Traditional storage devices and resources are
h
provided by different vendors. As a result, the resources cannot be shared. In the data center,
:
the storage resources are isolated from each other.
s
r ce
In a traditional storage system, centralized metadata management mechanism is adopted.
ou

This mechanism allows the metadata to record the distribution of the logical unit number
es
(LUN) data with different offsets on the hard disk. For example, the initial 4 KB of data in
R
LUN1+LBA1 is distributed on LBA2 of the 32nd hard disk. Each I/O operation initiates a query
n g
request to the metadata service. As the system scale grows, the metadata size also increases.
r ni
However, the concurrent operation capability of the system is subject to the capability of the
a
server running the metadata service. As a result, the metadata service eventually becomes
e
e L
the performance bottleneck of the system.
or
M
Distributed Server SAN Architecture
Server Server Server Server 1 Server 2 Server N
VM VM VM VM VM VM
VM VM VM VM VM VM
VBS VBS VBS
n
InfiniBand/10GE Network
/e
OSD OSD... OSD OSD OSD... OSD OSD OSD... OSD
o m
i.c
Cache Cache Cache
... ... ...
Local storage on the server Storage server Storage server Storage server
w e
ua
Shared storage resource pool Horizontal expansion of performance
h
and capacity
g.
n i n
ar Page 8
l e
: //
ttp
 Resource sharing within a data center: A huge storage resource pool can be created within a
h
data center to meet the requirements for storage capacity, performance, and reliability and
:
realize the resources sharing and unified management.
s
r ce
This is a new choice for storage of a cloud data center.
ou

es
VBS and OSD will be introduced later.

R
The DHT algorithm used by FusionStorage has the following features:
n g
ni
 Balance: Data is distributed to all nodes as evenly as possible, thereby balancing load
r
among all nodes.
e a Monotonicity: When new nodes are added to the system, system data is evenly
L

e
distributed to all nodes again. However, data migration is implemented only on new
or
nodes, and the data on the existing nodes is not significantly modified.
M  FusionStorage uses its distributed architecture to organize the dispersedly distributed, low-
efficiency SATA or SAS disks into an efficient storage pool. This pool is similar to a SAN but
provides higher I/O performance than SAN devices do, improving the storage performance by
one to three times.
 FusionStorage uses distributed stateless engines. These engines are deployed on each server,
thereby preventing performance bottlenecks that may be caused by centrally deployed
engines. Moreover, these distributed engines deployed on standalone servers consume less
CPU resources but provide much higher IOPS and throughput than centrally deployed engines
do.
 An example is provided as follows: A system has 20 servers that need to access the
storage resources provided by FusionStorage, and the bandwidth that each server
provides for the storage plane is 2 x 10 Gbit/s. One VBS module (storage engine) is
n
deployed on each server, so that the total throughput can reach 400 Gbit/s (20 x 2 x 10
/e
o
Gbit/s). With the growth of the cluster scale, storage engines can be linearly added, m
thereby eliminating the performance bottlenecks that may be caused by centralized
engines in traditional dual-controller or multi-controller storage systems.
e i.c
w
h ua
g.
n i n
ar
l e
: //
h ttp
s :
r ce
s ou
Re
n g
rni
e a
e L
or
M
Using FusionStorage in Two Application
Scenarios
Cloud resource pools Databases and key applications
Public Private Development Financial
VDI statements ERP Marketing CRM
cloud and test
cloud
n
OpenStack
HANA
/e
m
Virtualization platforms Query and search Data analysis Physical deployment
o
i.c
SCSI/iSCSI
e
FusionStorage
High-speed networks IB/GE/10GE
w
...x86 server
h ua
Distributed cache x86 server
DHT ring
g.
Tight consistency algorithm
Distributed cache
n i n
ar Page 10
l e
: //
ttp
 Compatibility: Compatible with mainstream databases, virtualization platforms, and servers.

: h
Converged deployment: Converged deployment of the virtualization platforms and database
s
ce
resource pools enables them to share one FusionStorage storage resource pool in the data
center.
r

s ou
FusionStorage uses SSDs to replace HDDs as high-speed storage devices and the InfiniBand
Re
network to replace GE or 10GE networks to provide higher bandwidth. Therefore,
g
FusionStorage is ideal for processing massive data in real time and meets the performance
n
ni
requirements.
a r
L e
e
or
M
When to Choose FusionStorage
Backup and Cloud resource
Integration of archiving of pools and OLAP
Core system blocks and files databases
massive files
n
Typical Scenario
/e
o m
i.c
Distributed
Product OceanStor 18000 OceanStor V3
Storage OceanStor 9000
Storage storage -
e
FusionStorage
w
ua
Restrictions 1. Object storage and file storage are not supported.
h
2. At least three nodes are required. Over 100 TB storage is
recommended to choose FusionStorage.
g.
n i n
ar Page 11
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
FusionStorage: Hardware and Software
Compatibility
 Hardware compatibility involves:
 Servers
 HDD disks, SSD disks, PCIe SSD cards or disks
n
/e
 RAID cards, Ethernet adapters, and Infiniband cards
 Software compatibility involves:
o m
i.c
 Virtualization platforms
 Operating systems
w e
ua
 Database software
. h
i n g
r n
l e a Page 12
: //
ttp
 For details about hardware compatibility, visit the following link:
: h
http://support.huawei.com/onlinetool/datums/fusioncloud/comptool/index.en.jsp
s

r ce
For more information about Huawei products compatibilities, visit the following link:
ou
 http://support-open.huawei.com/ready/index.jsf
es
R
n g
r ni
e a
e L
or
M
FusionStorage: Logical Architecture
VM 1 VM 2
FusionStorage FusionStorage Manager
Manager (active) (standby)
n
Server 1 Server 2 Server 3 Server 4 Server 5 Server 6
/e
FusionStorage FusionStorage FusionStorage FusionStorage FusionStorage FusionStorage
Agent Agent Agent Agent Agent Agent
m
MDC VBS MDC VBS MDC VBS VBS VBS
o
i.c
OSD OSD OSD OSD OSD OSD OSD OSD OSD OSD
e
Management&Compute Management&Comput Compute&Storage Compute&Storage
&Storage node Compute node Storage node
node node
w
e&Storage node
ua
 FusionStorage Manager (FSM) is the FusionStorage management module. It provides operation and maintenance functions such as
alarm reporting, monitoring, logging, and configuration. Generally, FSM nodes are deployed in active/standby mode.

. h
FusionStorage Agent (FSA) is an agent process deployed on each node to enable the communication between the current node and the
FSM node. An FSA includes MDC, VBS, and OSD processes. According to the different configuration requirements, different processes
are enabled on different nodes to realize specific functions.
i n g
r n
l e a Page 13
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
FusionStorage: Logical Architecture
 Metadata Controller (MDC): The metadata control component of FusionStorage for controlling
distributed cluster status, data distribution rules, and data rebuilding rules. The MDC processes are
deployed on ZooKeeper disks of three nodes to form the MDC cluster.
 Virtual Block System (VBS): The virtual block storage management component which manages volume
n
metadata and provides the distributed cluster access point service that enables computing resources
/e
to access distributed storage resources through VBS. Each node is deployed with one VBS process by
default, and all VBS processes form the VBS cluster. Multiple VBS processes can be deployed on one
node to improve the I/O performance.
o m
i.c
 Object-based Storage Device (OSD): Performs I/O operations. Multiple OSD processes are deployed on
each server. By default, one disk corresponds to one OSD process. When SSD cards are used as the
main storage, multiple OSD processes can be deployed on one SSD card to maximize the SSD card
w e
ua
usage and performance. For example, one 2.4 TB SSD card supports a maximum of six OSD processes,
h
each of which manages the I/O operations for 400 GB space.
g.
n i n
ar Page 14
l e
: //
ttp
 FusionStorage uses the data storage disks of the server.
: h
s
r ce
s ou
Re
n g
r ni
e a
e L
or
M
FusionStorage: Deployment Mode
 Converged deployment
 VBS and OSD are deployed on the same server.
 This mode is recommended for virtualization applications.

n
 Separate deployment
/e
VBS and OSD are deployed on different servers.
o m
i.c

e
 This mode is recommended for high-performance database
applications.
w
hua
g.
n i n
ar Page 15
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Contents
1. Product Overview

n
o m
i.c
w e
hua
g.
n i n
ar Page 16
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Basic Concepts (1)
 DHT: Distributed Hash Table, indicating the data routing algorithm for FusionStorage.
 Partition: A data partition represents a data area represented by a fixed Hash segment on the DHT ring.
 Key-Value: Data on the bottom-layer disks is divided into Key-Values. Each Key-Value represents a
data block.
n
Hash Partition Physical
/e
m
addressing Mapping the node
o
Key 1 physical space
Pn P1 Value 1
i.c
Key 2 Disk 1 Value 2
e
Key 3 ... P2
Value 3
DHT
w
ua
Key 4 Disk 2
Value 4
...
h
P6 P3
...
Key n
g.
Disk n
Value n
n
P5 P4
n i
ar Page 17
l e
: //
ttp
 The DHT ensures that the I/O operations performed by upper-layer applications are evenly
h
distributed on the hard disks of various servers and the load is globally balanced.
:
s
ce
 Decoupled from vendor-specific hardware: Hardware and software are highly integrated in
r
traditional storage systems, whereas server SAN products are decoupled from hardware and
not bound to specific hardware.
s ou

Re
Converging storage and computing: The server SAN is built on a universal x86 server,
g
converging both computing and storage.
n
r ni
e a
e L
or
M
Basic Concepts (2)
 Resource pool: a storage pool that consists of a set of partitions, corresponding to the
DHT ring.
 Volume: Application volume indicates an LBA continuous addressing visible to the

application.
n
Server 1 Server 2 Server 3
/e
Volume 1 Volume 2 Volume 3 Volume 10 Volume 11
o m
P1 P2 Px
e i.c
w
Resource pool 1 Resource pool 2
...
ua
P1 P2 Py
h
Disk Disk Disk Disk Disk
.
Disk
i n g
r n
l e a Page 18
: //
ttp
 A resource pool is similar to a RAID in the SAN. Compared with RAID, a resource pool has the
following advantages:
: h
s
ce
 A resource pool can contain a maximum of 96 disks (two copies), which provide ultra-
r
large storage space and prevent storage insufficiency on some frequently-accessed
disks.
s ou

Re
Dynamic hot backup: All the hard disks can function as the hot spare disks for resource
pools.
n g
ni
 Simple structure: Uses two-layer structure of resource pools and volumes, rather than
a r
LUN structure. The servers can directly connect to the volumes.

L e
The system automatically distributes data blocks on the hard disks of various servers. Data
e
or
that is frequently or seldom used is evenly distributed on the servers, thereby preventing
hotspots in the system.
M
Basic Concepts (3)
 Data copies: FusionStorage uses the multiple data replica mechanism to ensure data
reliability. That is, the same data can be replicated and stored as 2 or 3 copies.
Server 1 Server 2 Server 3

Disk 1 Disk 2 Disk 3
n
/e
P1 P2
P2’ P3 P4
P4’ P5 P6
P6’ P7 P8
P8’ P9 P10 P11 P12
m
P5 ’ P9’
P9 P17 ’ P21’
P21 P1 ’ P10’
P10 P13 ’ P22’
P22 P2 ’ P6 ’ P14 ’ P18 ’
o
Disk 4 Disk 5 Disk 6
e i.c
P13 P14
P14’ P15 P16
P16’
w
ua
P17 P18
P18’ P19 P20
P20 ’ P21 P22 P23 P24
P7 ’ P11’
P11 P19 ’ P23’
P23 P3 ’ P15 ’ P24’
P24 P4 ’ P8 ’ P16 ’ P20 ’
h
P12’
P12
g.
n i n
ar Page 19
l e
: //
ttp
 Each volume is divided into segments of 1 MB by default, and the segments are stored on
cluster nodes based on the DHT algorithm.
: h
s
ce
 P1' on disk 2 of server 2 is a copy of data block P1 on disk 1 of server 1. P1 and P1' are two
r
data copies of the same data block. If disk 1 becomes faulty, P1' can take the place of P1 to
provide storage services.
s ou

Re
The data fragment distribution algorithm ensures that active/standby copies are evenly
g
distributed to different hard disks of the servers. In this way, each hard disk contains the
n
ni
same number of active/standby copies.

a r
When a node is added or deleted due to a failure, FusionStorage employs the data rebuilding
L e
algorithm to balance load among all nodes after system rebuild.
e
or
M
FusionStorage: Data Routing Principles
Partition Disk
P1 Disk 1
Generate
(LUN1, LBA1) key 1 Key 1 = (LUN1, LBA1), Hash (Key 1) P2 Disk 2
, Data Data Pn P1
P3 Disk 3
... P2
Pn Disk N
OS VBS Hash space Routing query table
n
/e
P6 P3
Disk 3
P5 P4
OSD
o m
i.c
.
...
K2
K3
K5
e
K1
K4
.
.
w
Metadata region Data region
h ua
.
 The FusionStorage data routing uses layered processing mode.
 The VBS module determines the server hard disk where data should be stored after calculation.
i n g
The OSD module determines the specific hard disk position where data must be stored after calculation.
n


ar Page 20
l e
: //
ttp
 During system initialization, FusionStorage divides hash space (0 to 2^32) into N equal parts
h
based on the number of hard disks. Each part is called a partition. For example, in the two-
:
s
copy scenario, the system has 3600 partitions by default. If the system is equipped with 36
r ce
hard disks, each hard disk is allocated 100 partitions. The partition-hard disk mapping has
ou
been configured during system initialization and will be flexibly adjusted when the number of
s
hard disks in the system changes. The mapping table requires only small space, and
e
R
FusionStorage nodes store the mapping table in the memory for rapid routing purposes.
n g
When an application accesses FusionStorage, the SCSI command includes the LUN ID, LBA ID,
ni

r
and data to be read or written. The OS forwards the message to VBS on the node. Then VBS
a
creates a key using the LUN ID and LBA ID. The key contains the roundup result of the LBA ID
e
e L
divided by 1 MB. VBS uses the DHT hash algorithm to obtain an integer (ranging from 0 to
or
2^32) and directs the integer to a partition, locates the disk based on the "partition-hard
disk" mapping recorded in the memory, and forwards the I/O operation to the OSD to which
M the disk belongs.
 Each OSD manages one hard disk. During system initialization, OSD fragments disk space (1
MB for each fragment) for management and records the distribution information of each
fragment in the metadata management area. After receiving an I/O operation from VBS, OSD
locates the fragment where the requested data resides based on the key, obtains the data,
and returns the data to VBS, thereby completing the data routing process.
 An example is provided as follows: When an application needs to access the 4 KB data
identified by an address starting with LUN1+LBA1, FusionStorage first constructs
"key=LUN1+LBA1/1M", calculates the hash value for this key, performs modulo operation for
the value N, gets the partition numbers, and then obtains the hard disk of the data based on
the "partition-hard disk" mapping recorded in the memory.
n
/e
o m
e i.c
w
h ua
g.
n i n
ar
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Contents
1. Product Overview

n
o m
i.c
w e
hua
g.
n i n
ar Page 22
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
FusionStorage: VBS Module and I/O
Processing
 VBS is the access side of the storage function of FusionStorage. It has the following functions:
 Volume and snapshot management
 I/O access and processing
MDC AGENT VBS_CLI Kernel
n
/e
Heart VBM SCSI
Target
Beat Initiator
m
SCSI
Device Manager
o
VBP Block Driver
i.c
VSC
Client
DATANET
w e
ua
OSD OSD OSD
. h
g
VBM manages the volumes and snapshots, including volume creation, volume attaching, volume detaching, volume
i n
query, volume deletion, snapshot creation, snapshot deletion, and volume creation based on the snapshots.
r n
l e a Page 23
: //
ttp
 An I/O is processed by SCSI, VBP, and client in the VBS process.
: h
SCSI introduces the I/O from the kernel (VSC.KO) to VBS. After receiving an I/O request
s

ce
in the standard format of SCSI protocol, SCSI identifies a unique I/O using SCSI
r
quadruple (host_id, channel_id, target_id, and lun_id) and the offset and read-write
ou
data length (len) of the I/O on the block device, then delivers the I/O information to
s
e
Virtual Block Process (VBP).
R

n g
The VBP converts the I/O of the common format to “Key-Value” format used in
ni
FusionStorage, then delivers the I/O to the client. In this format, Key=tree_id (4
a r
Bytes)+block_no (4 Bytes)+ branch_id (2 Bytes)+snap_id (2 Bytes), where tree_id,
L e branch_id, and snap_id are the unique identifiers of volumes and snapshots in
e
FusionStorage, and block_no is the ID of the block where the I/O resides when the
or
volume is divided into 1 MB blocks.
M  After the I/O request enters the client, the client calculates the hush value based on
tree_id and branch_id in the Key to determine which OSD will process the I/O, then
sends the I/O to the OSD.
 OSD receives the I/O and returns the I/O layer by layer.
 The common attributes of the volumes and snapshots (such as sizes and names of volumes)
and some private attributes of the volumes and snapshots within the DSware (such as tree_id,
branch_id, and snap_id used to locate the storage location of volume and snapshot data in
the system) are saved in a private logical volume in the DSware. This private logical volume is
called a metadata volume.
FusionStorage: OSD Module and I/O
Processing
 Each disk managed by the
FusionStorage storage pool
corresponds to one OSD process.
OSD has the following functions:
n
 Disk management
/e
 I/O replication
I/O data cache processing
m

o
ei.c
w
hua
g.
n i n
ar Page 24
l e
: //
ttp
 OSD: works in active/standby mode. The MDC monitors the OSD status in real-time. When
h
the active OSD where a specified partition resides is faulty, the storage services will
:
s
automatically switch to the standby OSD, ensuring the service continuity.
RSM: realizes replication protocol.

r ce
ou

es
SNAP: manages the I/O functions of volumes and snapshots and the disk space.

R
CACHE: provides the cache function.
n g
ni
 AIO: distributes asynchronous I/Os to the underlying SMIO module and monitors media
r
failures over the SMIO interface.
e a
SMIO: distributes I/Os to physical media, monitors failures of physical media, and obtains the
L

e
disk information.
or  Each OSD manages one hard disk. During system initialization, the OSD fragments the disk
M space (1 MB for each fragment) for management and records the distribution information of
each fragment in the metadata management area. After receiving an I/O operation from VBS,
OSD locates the fragment where the requested data resides based on the key, obtains the
data, and returns the data to VBS. The data routing process is complete. When receiving a
write request, the active OSD queries the "partition-primary disk-secondary disk 1-secondary
disk 2" mapping and notifies each standby OSD of performing the write operation. Data is
returned to VBS only after both active and standby OSD nodes complete the write operation.
FusionStorage: Disk Partitions of OSD
Reserved area (1 MB)  Each 1 MB space of the disk is allocated to a

Metadata (4 MB) fixed key.
n
Key area  A certain number of consecutive keys form a
/e
Chunk 1 chunk.
Value area
 The storage space of a partition consists of
o m
i.c
...
Key area one or more chunks.
Value area
Chunk N
w e
VDB structure (blank disks)
h ua
g.
n i n
ar Page 25
l e
: //
ttp
 VDB
 Contact R&D for more information.

: h
s
r ce
s ou
Re
n g
r ni
e a
e L
or
M
VDB (Key-Value DB) VDB Name
Version
Reserve
Disk Size
Version Information Key Size
Reserved area (1 MB) Configuration Value Size
Chunk Size
Stat Information
Metadata (4 MB) Reserve
Partition Information
n
Used Space
Chunk Information
/e
Key area Key Num
Partition Number
Chunk 1
m
Free Flag Partition Offset
Value area
o
Key 1 Chunk Number
i.c
Chunk Offset
...
...
Free Flag Reserve
e
Key area Key N
Partition ID
Chunk N
w
Chunk Number
ua
Value area 1..M Reserve
Value 1
h
Value 2 Chunk ID
. Partition ID
...
g
Value N Begin Offset
n
1..K
i
End Offset
n
Reserve

ar Page 26
l e
: //
ttp
 Contact R&D for more information.
: h
s
r ce
s ou
Re
n g
r ni
e a
e L
or
M
FusionStorage: MDC Functions
 A Metadata Controller (MDC) is a highly reliable cluster. It employs the High Availability
(HA) mechanism to ensure the high availability and reliability of the system. The MDC
functions are as follows:
 ZooKeeper clusters are used to reliably save metadata such as the topology, OSD view, partition
view, and VBS view.
n
 The partition allocation algorithm is used to ensure the reliability of the RAID storing multiple
/e
data copies.
o m
i.c
 Obtains and sends notifications about the status changes of OSD and VBS nodes through
e
information exchange with the nodes.
w
 Realize capacity expansion and reduction, status query, and maintenance through information
ua
exchange with the Agent.
h
The heartbeat detection mechanism is used to monitor the status of OSD and VBS.
.

i n g
r n
l e a Page 27
: //
ttp
 To ensure the reliability of FusionStorage with typical deployment, one active ZooKeeper and
h
two standby ZooKeepers are deployed, and each management node is deployed with one
:
s
ZooKeeper process. The active and standby mode management is performed by the
ZooKeeper internal mechanism.
r ce

s ou
The MDC process and ZooKeeper process work in C/S mode and communicate with each
Re
other over TCP. The MDC can connect to any ZooKeeper server and maintain the TCP
connection. If the TCP connection breaks down, the MDC switches to another ZooKeeper
server.
n g
r ni
e a
e L
or
M
FusionStorage: MDC Functions
 A ZooKeeper (ZK) is a distributed service framework that solves problems in distributed applications,
such as uniform naming, status synchronization, cluster management, and distributed application
configuration management. The ZooKeeper functions are as follows:
 Active/standby MDC management: One active MDC and two standby MDCs are deployed. After the MDC
processes are started, all MDC processes register in the ZooKeeper, and the first registered MDC is the active MDC.
n
/e
When the MDC processes are running, the ZooKeeper records the active and standby MDCs information and
monitors the health of MDC processes through the heartbeat mechanism. Once the active MDC process is faulty,
m
another active MDC will be selected.
Data storage: When the MDC processes are running, information about various control views is generated,
o
i.c

including the target views, middle views, and I/O views. The information is saved, updated, queried, and deleted
e
using interfaces provided by the ZooKeeper.
w
 Data synchronization: The data is updated to the active ZooKeeper and automatically synchronized to the two
ua
standby ZooKeepers, ensuring real-time synchronization of the data on both active and standby ZooKeepers. In
case of the ZooKeeper active/standby switchover, services are not affected.
. h
i n g
r n
l e a Page 28
: //
ttp
 To ensure the reliability of FusionStorage with typical deployment, one active ZooKeeper and
h
two standby ZooKeepers are deployed, and each management node is deployed with one
:
s
ZooKeeper process. The active and standby mode management is performed by the
ZooKeeper internal mechanism.
r ce

s ou
The MDC process and ZooKeeper process work in C/S mode and communicate with each
Re
other over TCP. The MDC can connect to any ZooKeeper server and maintain the TCP
connection. If the TCP connection breaks down, the MDC switches to another ZooKeeper
server.
n g
r ni
e a
e L
or
M
FusionStorage: Views
OSD OSD Partition Partition Active OSD

OSD View
ID Status View ID OSD status
Standby OSD
OSD status
n
/e
Partition Active
I/O View View Standby OSD
ID OSD
m
Management OSD status
o
 I/O View: mapping between the partition and the active OSD.
e i.c
w
ua
 Partition View: mapping between the partition and active and standby OSD nodes. The
I/O view is a subset of the partition view.
. h
i n g
r n
l e a Page 29
: //

h ttp
The MDC obtains the OSD status through heartbeat messages. OSD reports specific
:
messages such as OSD capacity to the MDC per second. When the MDC fails to receives
s
ce
heartbeat messages from OSD for a specific time period (5s in this system), the MDC
r
considers OSD to be faulty, that is, OSD process disappears or the network between OSD
s ou
and MDC breaks down. In this case, the MDC sends messages indicating that the OSD is out
Re
of service, updates the OSD view, and sends each OSD a view change notification. OSD
g
determines the subsequent operation object based on the new view.
n

r ni
Multi-copy replication is dependent on the MDC views. In two-copy scenarios, when a
e a
client sends a write request to OSD, the OSD replicates the write request to the standby
e L
OSD of the same partition. In multi-copy scenarios, multiple copies of the write request are
or replicated to multiple standby OSDs.
M
FusionStorage: Interactions Between Main
Modules
MDC cluster
ZK disk
ZK disk ZK disk
ZooKeeper
ZooKeeper ZooKeeper Resource pool
MDC (Leader)
n
MDC MDC OSD OSD OSD
/e
Home OSD OSD
MDC OSD
OSD OSD
Obtain I/O view.
m
OSD
Resource pool
o
Report/notify status.
i.c
Inform status
OSD
change.
e
OSD OSD OSD Inform status changes.
Obtain I/O view. VBS cluster
w
OSD OSD
Query metadata.
ua
OSD OSD OSD
VBS (Leader) VBS
OSD Synchronize metadata.
. h
i n g
r n
l e a Page 30
: //
ttp
 When the system starts, MDCs interact with ZooKeepers to determine the leader MDC. The
h
leader MDC and other MDCs monitor each other and the leader MDC determines the MDC
:
s
that replaces a faulty MDC. If other MDCs find that the leader MDC is faulty, the MDCs
r ce
interact with the ZooKeepers to elect a new leader MDC.
s ou
When an OSD starts, it looks for its home MDC and reports its status to the home MDC. The
Re
home MDC sends the OSD's status changes to the connected VBS. If an OSD's home MDC
becomes faulty, the leader MDC specifies another MDC to replace the faulty one. A
n g
maximum of two resource pools can belong to the same MDC.
r ni
When a VBS starts, it looks for and registers with the leader MDC and queries from the leader
a

e
MDC whether it is the leader VBS. (The active MDC maintains a dynamic VBS list and
L
e
synchronizes the VBS list to other MDCs, so that the MDCs can inform VBSs with the OSD’s
or
status changes.) The VBS obtains I/O views from the leader MDC. The leader VBS obtains
M
metadata from OSDs while other VBSs get metadata from the leader VBS.
 Multiple VBS processes might be involved in a FusionStorage system. When multiple VBSs
perform operations on the metadata volume simultaneously, data might be damaged. To
avoid this, the VBS is deployed in active/standby mode, where only the active VBS can
operation on the metadata volume, whereas all standby VBSs are not allowed to do this.
FusionStorage has only one active VBS, and the MDC determines the active/standby role of
each VBS. The heartbeat messages between the VBSs and MDC ensure that at most one
active VBS is elected.
 Only the active VBS can operate on the metadata; therefore, the volume and snapshot
management commands received by standby VBSs must be forwarded to the active VBS. If
attachment and detachment processes are required, after completing operations on
metadata, the active VBS must forward the commands to the target VBS for volume
attachment and detachment.
n
/e
o m
e i.c
w
h ua
g.
n i n
ar
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
FusionStorage: Flexible Capacity Expansion
Computing network (10GE)

Server 1 Server 2 Server 3 Server 4 Server 5
App App App App App App
n
/e
To-be- To-be-
FusionStorage distributed integrated storage added added
m
node 1 node 2
o
i.c
Cache Cache Cache
SSD SSD SSD
HDD HDD HDD
w e
hua
g.
n i n
ar Page 32
l e
: //
ttp
 FusionStorage uses the distributed architecture to support easy capacity expansion and ultra-
large storage capacity.
: h
s
ce
 FusionStorage ensures rapid load balancing after capacity expansion and avoids migration of
a large amount of data.
r

s ou
FusionStorage supports flexible capacity expansion and allows both concurrent and separate
Re
expansion of computing nodes, hard disks, and storage nodes.
n g
FusionStorage evenly distributes caches and bandwidths of engines to all nodes, ensuring
ni
linear increase of the system IOPS, throughput, and cache as well as the number of nodes.
a r
L e
e
or
M
FusionStorage: Rapid Data Rebuild
Slice-based data Hardware Automated fault

storage fault detection
n
/e
o m
Parallel data restore Automatic data
ei.c
w
by multiple nodes replication
h ua
g.
n i n
ar Page 33
l e
: //
ttp
 Rebuilding 1 TB data takes less than 30 minutes.

: h
Each hard disk in the FusionStorage system stores multiple DHT partitions, whose data copies
s
ce
are scattered on other nodes in the system based on certain distribution policies. When
r
detecting a hard disk or server fault, FusionStorage automatically repairs data in the
background.
s ou

Re
The repair mechanism allows FusionStorage to simultaneously restore a minimal amount of
g
data on different nodes because the data copies are stored on different storage nodes. This
n
ni
mechanism prevents performance deterioration caused by restoration of a large amount of
r
data on a single node, and therefore minimizes adverse impacts on upper-layer services.
a
L e
e
or
M
FusionStorage: Cache Write Mechanism
 After receiving write I/Os from VBS, OSD stores the write I/Os in SSD cache. Then the data write to the local node is
complete.
 OSD periodically flushes write I/O data from SSD cache to disks in batches. The write cache has a watermark. The OSD
node will write data from the cache to HDDs if the amount of cached data exceeds the watermark even when the
flushing period has not arrived.
 FusionStorage supports passthrough of large blocks. By default, data blocks greater than 256 KB will be written directly
n
/e
to HDDs rather than being cached. This configuration can be modified.
OSD
o m
i.c
Step 1
e
Cache
Read Memory
w
Read Write SSD
h ua
HDD
Drain
g.
n
HDD HDD
n i
ar Page 34
l e
: //
ttp
 FusionStorage can use certain server memory as the read cache and NVDIMMs or SSDs as the
write cache.
: h
s
r ce
s ou
Re
n g
r ni
e a
e L
or
M
FusionStorage: Cache Read Mechanism
OSD
Step 1
Step 2
Step 3 Step 4
n
/e
Cache
Read Memory
o m
Read Write SSD
ei.c
w
Cache
ua
HDD
h
HDD HDD
g.
n i n
ar Page 35
l e
: //
ttp
 FusionStorage uses the multi-layer read cache mechanism. The first layer is memory cache
h
that uses the least recently used (LRU) mechanism to cache data.
:
s
ce
 The second layer is SSD cache that leverages the hotspot read mechanism. FusionStorage
r
collects statistics on the access frequency of read data, caches data whose access frequency
ou
has reached the threshold onto SSDs, and moves data that has not been accessed for a long
s
period of time out of SSDs.
Re

n g
FusionStorage supports the prefetch mechanism. It collects statistics on the relevancy of read
ni
data, automatically reads the data block that is highly relevant when reading a piece of data,
r
and caches the read data onto SSDs.
a

L e
After receiving a read I/O request from the VBS module, the OSD module performs the
e
or
following operations:
Step 1: The OSD module checks whether required I/O data exists in the memory read
M

cache. If yes, the OSD module responds to the VBS module and moves the I/O data to
the head of the LRU queue. If no, Step 2 is performed.
 Step 2: The OSD module checks whether required I/O data exists in the SSD read cache.
If yes, the OSD module responds to the VBS module and increases the access
frequency of the I/O data. If no, Step 3 is performed.
 Step 3: The OSD module checks whether required I/O data exists in the SSD write
cache. If yes, the OSD module responds to the VBS module and increases the access
frequency of the I/O data. If the access frequency of the I/O data reaches the threshold,
the OSD module caches the data in to SSD read cache. If no, Step 4 is performed.
 Step 4: The OSD module checks whether required I/O data exists in the HDDs. If yes,
the OSD module responds to the VBS module and increases the access frequency of
the I/O data. If the access frequency of the I/O data reaches the threshold, the OSD
module caches the data in to SSD read cache.
n
/e
m
 Read repair: When failing to read data, FusionStorage identifies the failure cause. If the data
o
i.c
cannot be read from a disk sector, FusionStorage retrieves the data from other copies of the
data on another node and writes the data back into the original disk sector. This mechanism
w
ensures the correct number of data copies and data consistency among data copies.e
h ua
g.
n i n
ar
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
FusionStorage: Distributed Cache
FusionStorage: distributed cache resource pool
Host Host Host Host
APP 1 APP 2 APP 1 APP 2

n
Cache resource pool Cache resource pool
/e
7.2 TB
> 9.6 TB
o m
i.c
Storage Storage Storage > Storage Storage Storage Storage
SRV2
e
SRV1 SRV3 SRV1 SRV2 SRV3 SRV4
w
ua
Cache sharing, allowing for scale-out
. h
i n g
r n
l e a Page 37
: //
ttp
 Each disk on FusionStorage servers uses independent cache and bandwidth, preventing a
h
large number of disks competing limited bandwidth between computing devices and storage
:
devices in a storage system.
s
r ce
FusionStorage can use certain server memory as the read cache and NVDIMMs or SSDs as the
ou

write cache. Caches are evenly distributed to all nodes. The total cache size on all servers is
es
far greater than that of external storage. Even when using large-capacity and low-cost SATA
R
disks, FusionStorage can still provide one to three times higher I/O performance.
n g
ni
 FusionStorage can use SSDs for caching data. In addition to providing high capacity and the
r
write cache function, the SSDs can collect statistics on and cache hot data, further improving
a
e
system performance.
L
e
or
 FusionStorage supports the cache disabling mode and a variety of cache media, such as
NVDIMMs, SSD cards, SSDs, and memory.
M
Contents
1. Product Overview

n
o m
i.c
w e
hua
g.
n i n
ar Page 38
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
FusionStorage: Read I/O Process
Server  An application delivers a read I/O request to the OS.
VBS APP The OS forwards the I/O request to the VBS module of
the local server. Then the VBS module uses the data
routing mechanism to determine the primary OSD
module where the data resides based on the LUN and
n
/e
LBA information in the read I/O. If the primary OSD
module malfunctions, the VBS module reads requested
OSD
Server data from the secondary OSD module.
o m
i.c
Primary  After receiving the read I/O request, the primary OSD
e
SSD HDD module obtains the requested data based on the read
w
cache mechanism and returns a data read success
ua
acknowledgement to the VBS module.
. h
i n g
r n
l e a Page 39
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
FusionStorage: Write I/O Process
Server
VBS APP
n
/e
o m
i.c
Server sync Server
e
OSD OSD
sync
w
Primary sync Secondary
ua
SSD async HDD SSD async HDD
. h
i n g
r n
l e a Page 40
: //
ttp
 An application delivers a write I/O request to the OS. The OS forwards the I/O request to the
h
VBS module of the local server. Then the VBS module uses the data routing mechanism to
:
s
determine the primary OSD module where the data resides based on the LUN and LBA
information in the write I/O.
r ce

s ou
After receiving the write I/O request, the primary OSD module simultaneously writes the data
Re
to both the SSD cache of the local server and the secondary OSD module of the other server.
The secondary OSD module simultaneously writes the data to the SSD cache of the local
n g
server. After receiving a data write success acknowledgement from the SSD cache of the local
r ni
server and the secondary OSD module, the primary OSD module returns a data write success
a
acknowledgement to the VBS module. In the meantime, data is flushed from the SSD cache
e
e L
to HDDs.
or
 If a piece of data is stored in three identical data copies, the primary OSD module
M
simultaneously writes the data to both the secondary OSD and third OSD modules.
FusionStorage: Data Processing Process
offset (1020k) + len (8k)
1020k 4k
Virtual disk |----tree id-------|---blockid-------|---bid--|---sid--|
0 1 2 0 0 0 1 0 0 0 0 0 0 0 0 KEY 1
VBS
0 0 0 1 0 0 0 1 0 0 0 0 KEY 2
key=KEY1, offset=1020k, len=4k
key=KEY2, offset=0k, len=4k
n
/e
OSD-client key DHT partition ioview Route target:
primary OSD
OSD (primary) DHT ptview

o m
i.c
key partition Route target:
secondary OSD
SNAP key Key

mgmt
w e
Data write location and
snapshot key processing
ua
OSD (secondary)
CACHE I/O scheduling, read hit, I/O sorting, and write
cache
AIO-DM I/O submission and disk fault and error

. h
g
management
DISK
n i n
ar Page 41
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Contents
1. Product Overview

n
o m
i.c
w e
hua
g.
n i n
ar Page 42
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
FusionStorage: Architecture
Storage
interface layer SCSI driver/iSCSI
Distributed linked clone

Backup
n
Storage management
Distributed snapshot
/e
m
Storage
o
service Distributed thin provisioning Disaster
i.c
layer
recovery
e
Distributed cache
w
ua
Storage Cluster status control Strong consistency
Cluster
h
replication protocol
.
engine self-
g
layer Distributed data routing Parallel data rebuild recovery
n i n
ar Page 43
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
FusionStorage Block Storage Function -
SCSI/iSCSI Block Interfaces
 The VBS node of FusionStorage provides block interfaces over SCSI or iSCSI:
 SCSI: SCSI is used when VBS is deployed on a physical server or VM (FusionSphere or KVM).
 iSCSI: VMware and Microsoft SQL Server clusters provide the storage access capability for the VM
and local server where no VBS is installed over iSCSI.
n
/e
CVM iSCSI-Target
VM VM VBS
m
VM VM
OSD
o
i.c
SCSI
VBS
w e
ua
OSD VMWARE- VMFS
ESXi
UVP/KVM
h
iSCSI-Initiator
Hardware
media
g.
Hardware media
n i n
ar Page 44
l e
: //
ttp
 If SCSI storage interfaces are used, the following functions are supported: snapshot, snapshot
backup, and linked clone.
: h
s
ce
 Users access storage devices by connecting the local initiator to the iSCSI target provided by
the VBS.
r

s ou
FusionStorage supports the following security access standards:

Re
CHAP identity authentication
n g
ni
 LUN Masking to authorize a host to access LUNs
a r
L e
e
or
M
FusionStorage: Thin Provisioning
2 TB 2 TB 2 TB Conventional
configuration
n
/e
m
300 GB 600 GB 900 GB
Smart thin provisioning
o
Time
e i.c
w
ua
 Thin provisioning remarkably improves storage utilization compared with the conventional method of
directly allocating physical storage resources.
. h
FusionStorage supports automatic thin provisioning. It delivers almost the same performance as
conventional SAN storage.
i n g
r n
l e a Page 45
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
FusionStorage: Snapshot
 The snapshot mechanism of FusionStorage stores the status of data on logic volumes at certain points in time for possible exporting
and restoration.
 FusionStorage uses the Redirect-On-Write (ROW) mechanism to create snapshots, which imposes no adverse impact on volume
performance.
 Infinite snapshots: Snapshot metadata is distributively stored and supports horizontal expansion, which eliminates bottlenecks
n
caused by centralized storage. There is no limit on the number of snapshots that can be created.
/e
 Quick volume restoration: Data migration is not required for disk restoration. It takes only 1 second to restore a volume using a
snapshot. However, if physical SAN storage is used, volume restoration takes several hours.
1
4
2
5
3
6
1
4
2 3
o m
i.c
5 6
7 8 9 7 8 9
1
4
2
5
3
6
1
4
2
5
3
6
11 12 13 11 12 13
w e
ua
7 8 9 7 8 9
1 2 3 1 1 1 1
. h
11 15 13
g
4 5 6 1 2 3 5 4 5 6
n
7 8 9
i
18 7 8 18
n
Initial volume First snapshot Data write Second snapshot Data write Equivalent

ar Page 46
l e
: //
ttp
 In the preceding figures, empty cells indicate that no physical disk space is allocated.
: h
s
r ce
s ou
Re
n g
r ni
e a
e L
or
M
FusionStorage: Linked Clone
 The linked clone function of FusionStorage
creates multiple clone volumes from one Parent
volume A
snapshot. Changes made to a clone volume
do not affect the snapshot and the other
clone volumes. Snapshot
C1
n
 A clone volume inherits all functions of a Clone volume
/e
B/Parent
common volume. You can create snapshots volume B
for a clone volume, use the snapshot to
m
Snapshot
Clone volume A
restore the data in the clone volume and C2
o
i.c
clone the data in the clone volume.
FusionStorage supports batch volume Clone volume Clone volume
e
 Clone
B1 B2 volume C
creation for VMs and can create hundreds of
volumes in a second.
w
ua
Clone volume Clone volumes as parent volumes Snapshots created for
 FusionStorage supports a linked clone rate of clone volumes
h
1:2048, effectively improving storage space
utilization.
g.
n i n
ar Page 47
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
FusionStorage: Custom Storage SLA
 A variety of resource pool service level agreement (SLA) specifications can be customized based on
service requirements.
 Data volumes of a VM can belong to different storage pools.
VM VM VM VM VM VM
n
/e
o m
e i.c
w
h ua
g.
n i n
ar Page 48
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Quiz
1. Explain server SAN and describe the relationship between
FusionStorage and server SAN.
2. Describe the logical architecture of FusionStorage as well as the

functions of FusionStorage Manager, MDC, VBS, and OSD.
n
/e
3. Explain FusionStorage deployment modes and application scenarios.
o m
ei.c
w
hua
g.
n i n
ar Page 49
l e
: //
ttp
1: Raise and answer questions onsite.
2: Raise and answer questions onsite.

: h
s
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Summary
 Basic Concepts and Key Principles
 Component Functions and Mutual Relationships
I/O Process Analysis

n

/e
o m
ei.c
w
hua
g.
n i n
ar Page 50
l e
: //
h ttp
s :
r ce
s ou
Re
n g
r ni
e a
e L
or
M
Thank You n
www.huawei.com /e
o m
ei.c
w
hua
g.
ni n
ar
l e
: //
http
s :
r ce
sou
Re
n g
rni
e a
e L
or
M
The privilege of HCNA/HCNP/HCIE:
With any Huawei Career Certification, you have the privilege on http://learning.huawei.com/en to enjoy:
n
 1、e-Learning Courses： Logon http://learning.huawei.com/en and enter Huawei Training/e-Learning
/e

o m
If you have the HCNA/HCNP certificate：You can access Huawei Career Certification and Basic Technology e-Learning
courses.
e i .c
If you have the HCIE certificate: You can access all the e-Learning courses which marked for HCIE Certification Users.
aw

Methods to get the HCIE e-Learning privilege : Please associate HCIE certificate information with your Huawei account, and
hu

email the account to [email protected] to apply for HCIE e-Learning privilege.

g .
 2、 Training Material Download
i n

arn
Content: Huawei product training material and Huawei career certification training material.

//le
Method：Logon http://learning.huawei.com/en and enter Huawei Training/Classroom Training ,then you can download
training material in the specific training introduction page.
p :
 3、 Priority to participate in Huawei Online Open Class (LVC)
t t

s :h
The Huawei career certification training and product training covering all ICT technical domains like R&S, UC&C, Security,
4、Learning Tools: rc e
Storage and so on, which are conducted by Huawei professional instructors.
u

s o
eNSP ：Simulate single Router&Switch device and large network.

R e
WLAN Planner ：Network planning tools for WLAN AP products.

n g
In addition, Huawei has built up Huawei Technical Forum which allows candidates to discuss technical issues with Huawei experts ,
ni
share exam experiences with others or be acquainted with Huawei Products.
a r
 Statement:
L e
r e
This material is for personal use only, and can not be used by any individual or organization for any commercial purposes.
o
M
HUAWEI TECHNOLOGIES CO., LTD. Huawei Confidential 1

HCIE-DC V1.0 Training Material 3 Cloud Data Center Computing and Storage Solutions

Uploaded by

Copyright:

Available Formats

HCIE-DC V1.0 Training Material 3 Cloud Data Center Computing and Storage Solutions

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

HCIE-DC V1.0 Training Material 3 Cloud Data Center Computing and Storage Solutions

Uploaded by

Copyright:

Available Formats

The privilege of HCNA/HCNP/HCIE:

email the account to [email protected] to apply for HCIE e-Learning privilege.

 Master design of modules of Huawei converged resource pool

3. Unified O&M Management

Resource FusionSphere Data Center 1

L e application approval process

 Internal cooperative parts

3. Unified O&M Management

Service Catalog Network Template

Data Center 1 Data Center 2

M policy management; O&M management; user management

Copyright © 2017 Huawei Technologies Co., Ltd. All rights reserved.

 X86 physical server

3. Unified O&M Management

Across regions and Support physical

Copyright © 2017 Huawei Technologies Co., Ltd. All rights reserved.

Virtual Virtual Virtual load Elastic IP User

1. Obtain information about

Cloud host VDC Cloud disk Elastic IP address

3rd device management

Infrastructure Storage Load

Servers, storage devices, network devices...

3. Unified O&M Management

network and increase the IT response speed.

PHP PHP Private IP address:

 Internal and external network isolation

2. What is the value of unified management and maintenance?

 Services provided in a VPC

Characteristics of the Converged Resource Pool Solution's unified

management and management

 Understand features of the ECS and physical host services.

2. Features of the ECS

3. Features of the Physical Host Service

6. Network Service Features

 When existing AZs of a VDC cannot meet service use requirements,

 When VDC organizational members are changed, a VDC service

A domain service manager can change the name of a VDC. u

Copyright © 2017 Huawei Technologies Co., Ltd. All rights reserved.

 Creating, modifying, and deleting the catalog.

 Creating an application service.

 Publishing an application service.

 Unpublishing an application service.

 Editing an application service.

 Deleting an application service.

 A VDC involves two user roles:

L e Resetting a user password.

or  Locking or unlocking a user.

 Applying for a service.

 Resetting the password of a user.

 Managing service instances a user has applied for.

2. Features of the ECS

3. Features of the Physical Host Service

6. Network Service Features

M  Cloning a ECS or converting a ECS into a VM template is not supported.

 Managing VMs on the existing virtualized platform is not supported.

2. Features of the ECS

3. Features of the Physical Host Service

6. Network Service Features

The feature is a test feature and strictly restricted.

 Restriction on automatic deployment of physical machines in the current version: