1. Power of the cloud, Microsoft working on cloud-native Teams Xiaogang Xin

2. About me Xiaogang Xin Principal Engineering Manager Microsoft ee.msup.com.cn Continuous Integration at Scale on Kubernetes Model and Operate Datacenter by Kubernetes at eBay

3. Agenda • M365的发展和背后支撑的几代基础架构 • Teams on K8S • 溯源云原生 • 微服务香吗？ • 快 = 灾难 • 社区是不是yyds？ ee.msup.com.cn

4. Microsoft 365 ee.msup.com.cn

5. Exchange Online deployment model ee.msup.com.cn

6. Exchange Online deployment model ee.msup.com.cn

7. Shared infrastructure - COSMIC App A App B App C Application Logic Application Logic Application Logic Security & Compliance Security & Compliance Security & Compliance Deploy & Validate Deploy & Validate Deploy & Validate Monitoring Monitoring Monitoring Capacity Management Capacity Management Capacity Management HA / DR HA / DR HA / DR Multi Clouds Multi Clouds Multi Clouds Auto-Scaling Auto-Scaling Auto-Scaling Bin-packing Bin-packing Bin-packing k8s k8s AKS k8s AKS k8s COSMIC ee.msup.com.cn k8s k8s

8. Teams on K8S COSMIC PARTNERS “I just want to run my application and not worry about the rest!” • Application Development • Deployment Management (via ADO RM – per your business cadence) • Scenario Monitoring & Management COSMIC PLATFORM “You want Kubernetes at scale? Here’s a namespace! Have fun!” • Global Presence – 100s of ephemeral clusters, 1000s of nodes – wherever Office customers need us (including sovereigns and air- gapped clouds) • Centralized Security, Compliance, Monitoring, Capacity Management, Patching, Rings, SDP, Workload placement, Bin Packing, COGS efficiency • Unified HA/DR, Routing - (For some scenarios) AZURE KUBERNETES ee.msup.com.cn “Here’s a cluster!” • Managed, Single-Cluster Control Planes • One-Touch Kubernetes Version Upgrades • Zero-Day Kubernetes Security Updates • Click-Button Cluster Scale-Up, Azure Policies, Azure Monitoring

9. Teams on K8S – Key learnings Running well in old environment Modern and Cloud Native • ee.msup.com.cn Machine centric --> Application centric

10. Teams on K8S – Key learnings Mercedes? ee.msup.com.cn

11. Let's re-think about Cloud Native ee.msup.com.cn

12. Let's re- think about Cloud Native ee.msup.com.cn • Cloud native technologies empower organizations to build and run scalable applications in modern, dynamic environments such as public, private, and hybrid clouds. Containers, service meshes, microservices, immutable infrastructure, and declarative APIs exemplify this approach. • These techniques enable loosely coupled systems that are resilient, manageable, and observable. Combined with robust automation, they allow engineers to make high- impact changes frequently and predictably with minimal toil.

13. What's NOT Cloud Native? Please think about these questions • What do we really want to achieve? • Why can't we build great products faster before? • Others are doing it, so we are afraid to miss it unless we start doing it?? • What cloud native means to your team and what is the real problem to be solved? ee.msup.com.cn

14. Let's see below microservice app - The Butterfly Effect token Client token Auth API API S/F TTL token TTL Auth failure Token cache expired Auth API latency increased 1 CPU overheat Query performance impacted Retry Retry Retry Query storm ee.msup.com.cn

15. The danger of "fast" - 3 cases A production cluster disappeared during Chinese New Year ! I want my namespace back ! All IPs were reclaimed ! A disaster caused by a left join ee.msup.com.cn Circuit Breaker

16. The danger of "declarative" • Kubernetes is nothing but a set of controllers based on declarative model • Controller loop will generate huge pressure to downstream sys ee.msup.com.cn

17. Always trust opensource community? ee.msup.com.cn

18. Always trust opensource community? ee.msup.com.cn

19. ee.msup.com.cn

20. ee.msup.com.cn We are hiring! 100+ openings in Suzhou/Beijing, all levels including engineer and manager xiaogangxin@microsoft.com

21. 关注msup公众号 获取更多工程效能实践案例