Computer Networks

Introduction
Week 2-Lecture 1

Introduction 1-1
Today’s lecture: Basic questions

What do we mean by “computer networks”?

What do computer networks do?

What do computer networks look like?
Why study computer networks?
What is a computer network?

A set of network elements connected together, that implement a set

of
protocols for the purpose of sharing resources at the end hosts
• Three important components:
• Core infrastructure:
• A set of network elements connected together
• Protocols:
• Needed to use the network
• Purpose:
• Sharing resources at the end hosts (computing devices)
What is a computer network?

A set of network elements connected together, that implement a set

of
protocols for the purpose of sharing resources at the end hosts
What do computer networks do?

A computer network delivers data between the end points/hosts

• One and only one task: Delivering the data

• This delivery is done by:

• Chopping the data into packets
• Sending individual packets across the network
• Reconstructing the data at the end points

Evolution of three components of computer networks!

• Infrastructure, protocols, purpose
Data delivery as a fundamental goal

• Support the logical equivalence of Interprocess Communication (IPC)

• Mechanism for “processes on the same host” to exchange messages

• Computer networks allow “processes on two different hosts” to

exchange messages

• Clean separation of concerns

• Computer networks deliver data
• Applications running on end hosts decide what to do with the data

• Keeps networks simple, general and application-agnostic

What do computer networks look like?

Three Basic components

• End hosts: they send/receive data

• Switches/Routers: they forward data

• Links: connect end hosts to switches, and switches to each other

What do computer networks look like?

End hosts, switches/routers, links

Why study computer networks?
#1: Has transformed and more importantly, is transforming
everything!
• Industry: core to and creator of many large and influential companies
• Google, Facebook, Apple, Cisco, Juniper, Akamai
• Communication
• Email, messenger, phones, VoIP, …
• Travel
• AirBnB, Uber, Maps, …
• Health
• Digital health, remote diagnostics, ….
• Entertainment
• Netflix, Prime
• Relationships
• Facebook, Instagram, Snapchat, the list is endless…
Why study computer networks?
#2: To learn how to design for scale!

• Tremendous scale
• 51% of world population
• 1.24 trillion unique web pages
• Every second, approximately
• > 2 million emails
• > 40000 Google search queries
• > 6000 Tweets

• Introduced the phrase “Internet Scale”

Why study computer networks?
#3: To learn how to design for diversity!

• Communication latency: Microseconds to seconds

• Bandwidth: 1Kilobits/second to 100Gigabits/second
• Packet Loss: 0-90%
• Technology: Wireless, satellite, optical, copper, …
• End hosts: Sensors, cell phones, computers, servers, datacenters, …
• Applications: www, voice, video, gaming, remote medicine
• Trust models: selfish (users), malicious (attackers), greedy (companies),
…
And yet, everything needs to work in tandem!
Why study computer networks?
#4: To learn how to design for
evolution!

1970 Today
Bandwidth 50 kbps 100+ Gbps

#End hosts < 100 computers 8 billion +

Telnet and
Applications File transfer
!!
Architectural principles, design goals and performance objectives in wired
networks

• What tasks get done?

• What is delivered (packets, files, …)?
• What are the semantics (reliability, ordering, …)?

• Where do tasks get done?

• At the network elements? At the end-hosts?
• How do end hosts interface with network elements?
• How do different network elements interface with each other?

• How tasks get done?

• What protocols and algorithms do each of these use?
• How to achieve various performance objectives (latency,
etc.)?
 A closer look at network structure:
• network edge: mobile network
• hosts: clients and servers
• servers often in data centers global ISP

home
 access networks, network
regional ISP
physical media: wired,
wireless communication
links
 network core:
 interconnected routers
 network of networks institutional
network

Introduction 1-14
Chapter 1: roadmap
1.1 what is the Internet?
1.2 network edge
 end systems, access networks, links
1.3 network core
 packet switching, circuit switching, network structure
1.4 delay, loss, throughput in networks
1.5 protocol layers, service models
1.6 networks under attack: security
1.7 history

Introduction 1-15
The network core
• mesh of interconnected
routers
• packet-switching: hosts
break application-layer
messages into packets
• forward packets from one
router to the next, across
links on path from source
to destination
• each packet transmitted at
full link capacity

Introduction 1-16
Many mechanisms: What do we mean by …

• Locating a destination? Naming, Addressing

• Finding path to the destination  Routing
• Sending data to the destination  Forwarding
• Failures, reliability,etc.. Distributed routing and congestion control
A computer network can be abstractly represented as a graph

Source Path

Path

Source

Destination

Destination
1. What does network sharing mean?
2. What are the performance metrics?
3. What are the various mechanisms for sharing networks?
What does network sharing mean?
The problem of sharing networks

• Must support many “users” at the same time

• Each user wants to use the network (send and receive data)
Limited resources

• Fundamental question:

How does network decide which resource to allocate to which user

at any given point of time?
What are the performance metrics?
What do we look at before commuting to
places?
• Which are the possible routes to take from DAIICT to home?
• Which one is the best?
• How long does it take to reach my home from DAIICT?
• What is the capacity of a road(s) between DAIICT and home?
Capacity
• How wide is the
road?
• How fat is the tunnel?
• How many cars can fit
at a time?
• One, two, three…?
 Performance metrics in computer networks!
Capacity is Bandwidth: Number of bits sent per unit time (bits per second, or
bps)
• Depends on

• Hardware Bandwidth
• Network traffic conditions

Time taken is Propagation delay: Time for one bit to move through the link
(seconds)
• Depends on
Hardware
•
• Network traffic conditions Propagation
Delay
• How large is the unit?
• Each bit is a pulse of some width.
• For example, each bit on a
• 1-Mbps link is 1 µs wide
• 2-Mbps link is 0.5 µs wide,

• The narrower each bit can become, the higher the bandwidth.
• This means more bits can get inside the tunnel
• So MORE DATA CAN FLOW WITHIN A TIME
Bandwidth-delay product (BDP)
Number of bits “in flight” at any point of time (bits)
• Bits sent, but not received

Bandwidth Bandwidth x delay

Propagation
• Same city over a slow link Delay
• Bandwidth: ~100Mbps
• propagation delay: ~0.1ms
• BDP = 10,000 bits (1.25KBytes)
• Between cities over fast link:
• Bandwidth: ~10Gbps
• propagation delay: ~10ms
• BDP = 100,000,000 bits (12.5MBytes)
What are the various mechanisms for sharing networks?
 How would you design a sharing mechanism?

Hint:
Think about sharing any
resource (say, a computer)
Two approaches to sharing networks

• Reservations
• On demand
Two approaches to sharing networks

• First: Reservations
• Reserve bandwidth needed in advance
• Set up circuits and send data over that circuit
• No need for packets
• Must reserve for peak bandwidth

• Peak bandwidth?
• Applications may generate data at rate varying over time
• 100MB in first second
• 10MB in second second …
• Reservations must be made for “peak”
Circuit switching: Implementing reservations since …

Telephone networks

• One of the many approaches to implementing reservations

• Mechanism:
• Source sends a reservation request for peak demand to destination
• Switches/routers establish a “circuit”
• Source sends data
• Source sends a “teardown circuit”
message
Circuit switching
end-end resources allocated to,
reserved for “call” between
source & dest:
• In diagram, each link has four circuits.
• call gets 2nd circuit in top link and
1st circuit in right link.
• dedicated resources: no sharing
• circuit-like (guaranteed)
performance
• circuit segment idle if not used by call
(no sharing)
• Commonly used in traditional
telephone networks

Introduction 1-34
Challenges with Circuit switching (reservation)

• Handling failures
• Resource underutilization
• Blocked connections
• Connection set up overheads
Fun Quiz
Google celebrates its birthday on September 27, although no one really knows the exact date when it was founded. Started
by two Stanford college friends, Larry Page and Sergey Brin, in 1998, it is a multi-billion dollar enterprise now. The name
comes from a simple misunderstanding when they were searching for another, actual, word that existed in academia and
meant a particular number. What word is that and what number does it denote?

Googol, 1 followed by 100 zeros

 Gmail was launched by Google on April 1, 2004, which led many to believe it was an April Fool’s joke. Before this service
the term ‘G-mail’ already existed from as early as 1998. This was used online by fans of a certain fictional obese cat, and
the original G-mail was known as “e-mail with cattitude”. What does the G stand for in the original G-mail?

Garfield

When Page and Brin built the first server rack for Google at Stanford, they were looking for a cabinet to house it that
was easy to assemble and disassemble. The server contained ten 4GB hard disks and two cooling fans. What colourful
and bountifully found system did they use to build the server stack?
Lego bricks

Who is the father of the Internet?

Vincent Cerf
 Solution: Packet switching

• Break data into smaller pieces

• Packets!

• Transmit the packets without any reservations

• And, hope for the best
Recap: Packet switching summary

• Goods:
• Easier to handle failures
• No resource underutilization
• A source can send more if others don’t use resources
• No blocked connection problem
• No per-connection state
• No set-up cost

• Not-so-goods:
• Unpredictable performance
• High latency
• Packet header overhead
Recap: Deep dive into one link: packet delay/latency

• Consists of six components

• Link properties:
• Transmission delay
• Propagation delay
OS internals:
•
• Processing delay
• Queueing delay
Traffic matrix and switch
•
internals:
• Processing delay
• Queueing delay
• First, consider transmission, propagation delays
• Queueing delay and processing delays later in the
course