Keynote: The Big Leap: Conquering the Legacy System and Building a Scalable Cloud Infra...
summary
A custom Kubernetes-based platform called Elastic Cloud runs ~50,000 pods on 3,000 servers and ~50,000 pods and aims to migrate legacy services without downtime by running physical machines and containers in parallel. Problems addressed include low resource utilization, slow service startups (default 140s), and weak CPU/network/disk isolation. Fixes include local storage, a custom scheduler with request-ratio classification and weighting, and sealed large containers; outcomes: CPU utilization from 25% to 40% with stddev 14→10 and a drastic drop in 99th‑percentile startup latency.
AI-written summary · about
chapters (9)
- 0:00Legacy system transformation
- 1:48Three years zero to one
- 3:35System integration challenges
- 5:22International scalability demands
- 7:55Performance optimization
- 9:01Orchestration algorithm
- 10:51Hybrid deployment stability
- 11:55Key insights and outlook
- 13:43Community recommendations
related talks
14mKeynote: The Big Leap: Conquering the Legacy System and Building a Scalable Cloud...
18mKeynote: Mind the Gap: Bridging Cloud Native Innovation with Real-World Use
22mKeynote: Cloud Native at Alibaba Scale
30mMigrating a monolith to Cloud-Native and the stumbling blocks
26mKeynote: A Platform for the Enterprise: Where Maturity and Innovation Intersect
03mKeynote: Time to Unlock the Edge Potential at Scale
25mKeynote: Infrastructure - Introduction
13mSponsored Keynote: Multi What???? Multi-cloud, Multi-cluster, Multi-tenanted K8s