Weathering the Cloud Storm: Building Resilient Geo-Distributed Apps

on YouTube ↗50m

" summary

A Slack-like messenger is deployed to Google Cloud using Spring Cloud, two Kubernetes clusters (US East and US Central), a distributed config server, attachments in Google Cloud Storage, and YugabyteDB for transactional storage. Demo shows multi-cluster Ingress / global load balancer routing a single IP to the closest cluster; deleting the East messenger deployment immediately shifts traffic to Central. Notes: AZ latency ~5 ms, cross-region ~65 ms, cloud majors average a major outage every 50 days (≈4 hours to recover); Yugabyte cluster fails over in ~3 seconds with zero RPO options.

AI-written summary · about

" chapters (23)

1. 0:00Introduction to geo-distributed apps
2. 1:03Defining geo-distributed applications
3. 3:12Cloud outages and resilience
4. 7:47Deployment strategies for high availability
5. 9:12Balancing latency and performance
6. 12:40Regulatory compliance in deployments
7. 15:28Building blocks of geo-distributed apps
8. 17:34Spring Cloud for distributed systems
9. 18:57Demo setup with Kubernetes clusters
10. 21:06Deploying application instances
11. 24:17Testing the multi-region deployment
12. 27:46Deploying in US Central region
13. 29:12Global load balancing explained
14. 31:17Testing load balancer routing
15. 34:11Database resilience strategies
16. 37:59YugabyteDB for distributed SQL
17. 39:24Single multi-region cluster setup
18. 41:06Read replicas for distant users
19. 43:10Dedicated regional clusters
20. 44:35Geo-partitioned cluster approach
21. 46:22Fault tolerance test demonstration
22. 48:09Scaling from single region
23. 49:14Further learning resources

" related talks

• 39mHow Geo-Distributed Apps Use Global Cloud Infra for Speed and ComplianceDenis Magda · CloudOpen @ Open Source Summit Europe 2022
• 31mToo Far Too Furious: Multi-Region Cloud ApplicationsDenis Magda · Scale By the Bay 2023
• 38mHow Geo-Distributed Apps Use Global Cloud Infra for Speed and ComplianceDenis Magda · CloudOpen @ OSS LATAM 2022
• 31mAn Adventure in Multi Cloud with Kubernetes and Distributed DataLeon Kuperman · Distributed SQL Summit Asia 2021
• 32mZero-Touch Fault-Tolerance for Cloud-Native Geo-Distributed Databa...Selvi Kadirvel, Aman Nijhawan · Data on Kubernetes Day North America 2023
• 44mToo Far, Too Furious: Multi-Region Spring Cloud ApplicationsDenis Magda · SpringOne at VMware Explore 2023 Las Vegas
• 26mA Winning Multi Cloud Data StrategyChirag Narang · Distributed SQL Summit Asia 2021
• 51mAchieving Resiliency With A Multi-Cloud Or Multi-Region ApproachDaShaun Carter · Philly ETE 2023