Some Lessons We Learned from Moving E-business Giant to Cloud Native
summary
Alibaba migrates core e‑business to Kubernetes (ACK) with an ELISA clouds layer, running ~1M containers across 100+ datacenters and ~10 large clusters. Engineers build a batch allocation planner to place hundreds of thousands of pods deterministically, a sitecar/sidecar injection system to manage auxiliary containers separately, and an in-place update mechanism so upgrades avoid full pod recreation. These changes reduce scheduler pressure, preserve layer caching, improve release predictability for peak events like Nov 11, and are being open sourced.
AI-written summary · about
chapters (16)
- 0:00Intro
- 0:38Alibaba's Journey to Cloud Native
- 2:23Current State
- 3:19Architecture
- 10:23Eliminate "Rich Container"
- 15:41The Biggest Benefit is Upgrade of Operating Model
- 18:15Cloud Native App Mgmt
- 23:28Workload Mgmt in Web-scale is Challenging
- 24:39Batch placement
- 29:32Application upgrade All default workloads recreate Pods during rolling update. Strategies they offered
- 30:02Introduce: in-place update
- 32:13Comparation
- 35:25Sidecar management
- 38:40Sidecar Set injection
- 39:38What can Sidecar Set do
- 41:00OpenKruise
related talks
44mSome Lessons We Learned from Moving E-business Giant to Cloud Native
30mMigrating a monolith to Cloud-Native and the stumbling blocks
38mThe Good, the Bad & the Ugly of Migrating Hundreds of Legacy Applications to Kubernetes
39mEmbracing Cloud Native at a Thriving, Established Company
32mKubernetes Namespaces Unleashed: Unlocking the Full Potential of You...
29mEvolve your web application for scale
29mFrom pets to cattle, the way of Kubernetes
17mKeynote: Extending Kubernetes: Application Orchestration to 1000’s of Clusters in...