Real-time Kubernetes: How Entain Australia 10x'd throughput with Linkerd
Talkat KubeCon + CloudNativeCon North America 2021 · on Friday, 2021-10-15· at 15:25· Concourse Hall 151 + Online·
summary
Entain Australia runs ~300 microservices across five clusters (≈3,000 pods) processing ~10 billion external updates per month and hits hot paths with gRPC connection reuse; installing Linkerd as a sidecar mesh converts connection-level pooling into request-level load balancing, yielding about 10x throughput during a bulk-load incident, immediate CPU and network smoothing, and lower inter-zone bills. Deployment uses Spinnaker to add annotations (no code changes), integrates Jaeger tracing and mTLS, and proves effective in production.
AI-written summary · about
chapters (22)
- 0:00Intro
- 0:11About Entain Australia
- 0:37Steve Gray
- 1:35Our Software Stack
- 1:54Our Platform is Cloud Native
- 2:15Some Key Metrics
- 2:36It Wasn't Always This Way....
- 3:01Our Microservices Journey
- 6:12The Challenges of Growth
- 6:37Service Discovery Evolution
- 7:08GRPC Woes....
- 7:52Rolling Out Code.....
- 9:04C'est La Vie?
- 9:33It Went Poorly
- 10:09Days of Effort Lost
- 10:45Solution Selection
- 11:43Our Selection
- 12:29The Deployment
- 14:17How? What? How is it....
- 15:43Unexpected Gains
- 16:25Other Features We Got....
- 17:18Our Journey Onward
related talks
01h 03mLinkerd Salon
39mService Mesh: There and Back Again
37mLinkerd End User Panel: Case Studies from Production
01h 10mService Mesh Introduction
35mDeep Dive: Linkerd
19mReplatforming a $4B Retailer Onto Kubernetes and Linkerd
12mService Mesh & Kubernetes Multicluster Communication With Linkerd
32mLinkerd and the Service Mesh