Smruti Prakash Mohanty

Disaster recocery and backup are boring but critical.

5

What is the reality check behind 99.999 SLA promised by hyperscalers ? UniSuper, one of Australia's largest retirement funds, faced a catastrophic outage on Google Cloud. Despite assurances of a 99.9999% SLA and deploying 1900 VMs, databases, and applications across two geo-redundant regions, an internal error tied to subscription management led to a complete service and data loss for days. Recovery was only possible many days later through backups from another service provider. Lessons Learned: 1. Limits of Geo-Redundancy:Even with high SLA promises, dual-region strategies tied to a single subscription can fail, so two subscriptions for prod which is not the common might be a mitigation and infra teams would def. not favor this. 2. The Case for Diversification: multi-cloud strategies might provide more reliable safeguards against catastrophic failures esp for backups. #CloudComputing #DataSecurity #BusinessContinuity

2

Next time you think backups are not important, just remember the time when Google Cloud accidentally deleted an entire cloud environment for an investment company managing over $80 billion. Luckily, they had extra backups with another cloud provider. https://lnkd.in/ej_H5rr2 #gitops #devops #devsecops #devopsengineer

4

New IRAP assessment out including new services. Many Australian customers are looking to experiment with how generative AI applications can help them better serve the Australian public. Customers can use two of the newly assessed services—Amazon Bedrock and Amazon DataZone—to help align with their governance, sovereignty, and security requirements up to the PROTECTED level.

14

If you missed the AWS Sydney Summit, here is the recording of our talk on using Generative AI with Kafka and Flink. Abhaya Chauhan https://lnkd.in/gAC6_Syt #kafka #flink #msk #aws #genAI

39

2 Comments

Doesn't matter the scale, mistakes do happen. As solutions architects, we need to always design for failure and never trust a single entity blindly, even if it's Google. Takeaway from this incident: It is better to have another backup with an alternative service provider along with the primary backup & save yourself from an outage. #cloudcomputing #googlecloud #aws #gcp #cloud #technology

8

🚨 UPDATE 🚨 Cloud Resume API Challenge Prize Pool Has Increased ...AGAIN! Rishab Kumar, REXTECH Community and I want to express our deepest gratitude to Damien Burks for being offering to donate to the Cloud Resume API Challenge by adding $200 to the overall prize value which is now worth $1200!! 𝗪𝗵𝗮𝘁'𝘀 𝗡𝗲𝘄? Prize pool increased to $1200 USD More opportunities to win valuable cloud certifications and credits 𝗖𝗵𝗮𝗹𝗹𝗲𝗻𝗴𝗲 𝗗𝗲𝘁𝗮𝗶𝗹𝘀: Check out the previous post for details on The Cloud Resume API Challenge —> https://lnkd.in/eB3YZm-f 𝗛𝗼𝘄 𝘁𝗼 𝗣𝗮𝗿𝘁𝗶𝗰𝗶𝗽𝗮𝘁𝗲: - Let us know if you’ll be taking part ✅ - Follow the challenge guidelines ✅ - Document your progress ✅ - Submit your completed project with a PR by July 31st ✅ 𝗥𝗲𝘀𝗼𝘂𝗿𝗰𝗲𝘀: YouTube Guides: - Cloud Resume API Challenge - https://lnkd.in/g7qptvgF - AWS Cloud Resume API Challenge - https://lnkd.in/gm7R7cA6 - GCP Cloud Resume API Challenge - https://lnkd.in/gZH9ySnd Documentation Guide - https://cloudresumeapi.dev ♻️ Share this post to help others boost their cloud careers! #cloud #cloudcomputing #cloudengineer #cloudchallenge #cloudresumeapichallenge

78

22 Comments

scaling pods and nodes in EKS is a real play, those who run EKS clusters, pods and scaling really knows their challenges and limitations. Gigi Sayfan has articulated challenges and workaround for some of it. learn about kubernetes and it scaling methods. learn about eks scaling, Pod auto scaling, ASG and its limitation on Pod tables. #kubernetes, #amazoneks, #awscommunity, ##awscommunitybuilders https://lnkd.in/gCHq6T4H

7

My new article on using Java's EpsilonGC to understand memory allocation in code without using a profiler. https://lnkd.in/g9wDTpdw #java #memoryprofiling #performanceengineering #performance

11

I'm not sharing this to bash one cloud provider or the other. I'm purely sharing this for educational and awareness purposes since such incidents can happen in any environment (any cloud and any data center). Safeguard your data, have backups not just of your core data, but your code and configuration as well. Test it frequently for security, relevance, history, RTO, and even your operational knowledge of team members. Test your red button preparedness frequently. 99.99% SLA means 52 minutes and 9.8 seconds of downtime per year 99.999% SLA means less than six minutes of downtime over the course of a year But, just using one service that is 99.999% up doesn't mean your whole architecture is that much resilient. Your architecture is only as resilient as your weakest link. Have a DR plan, test it, and implement it.

16

4 Comments

If you're using SQS FIFO take a look at these new metrics. They give you a lot more visibility into the behavior of your publishers.

15

this is a very good article which describes "Distributed Cache". https://lnkd.in/gfyHSMt4 #cache #distributed cache #systemdesing

1

Oops.. underlines the importance of type of data and applications that should go on the cloud providers in addition to having diverse multi-layer backup strategies.

26

6 Comments

Amped for #AWSreInvent? Same. We’ll be there in force – learning, mingling and sharing ideas. We’ll also be presenting some of the breakthrough work we’re doing around how our data and cloud environments help us to innovate with #AI . Are you going to Vegas? Drop a comment below. https://bit.ly/3OfBLfJ

4

1 Comment

Containerization is a powerful tool for managing workloads. However, not all workloads are suitable for containerization. In this tech talk I cover out how to handle virtualized workloads that don’t make sense to containerize while leveraging the robust self-healing capabilities of Kubernetes.

9

2 Comments

Creating Validating Admission Controller in Kubernetes using Go #WorkSmartWithK8s #kubernetes #admissioncontroller #golang #example https://lnkd.in/g7qqbjn6

9

Any modernization is incomplete if data layer architecture is not modernised. Good reference to support the argument.

11

On my way to #reinvent for the 4th time to compete in the AWS DeepRacer Grand Finals. Managed to finish 10th globally this year, so we will see how we go next week! There will be a live stream for my racing Tuesday morning PST, Wednesday lunchtime NZDT. #aws #deepracer #machinelearning #ai

64

9 Comments