Panel Discussion: Ask Us Anything: Microservices and Service Mesh
summary
Panelists compare Istio and Envoy, show concrete patterns: enable sidecar Envoy proxy to get immediate telemetry and mTLS, use context propagation for distributed tracing (requires near 100% rollout), and run canary rollouts and retries from the mesh. Envoy implements ring hash and maglev for sticky/partitioned routing and already proxies Redis at Lyft; replacing a giant edge proxy gave a ~10x scale gain. Expect control-plane costs, per-pod sidecar overhead, and tooling needs — Lyft spends 4–5 engineers operating Envoy; a thin 100–200 line envoy client simplifies adoption.
AI-written summary · about
chapters (13)
- 0:00Intro
- 1:12Service Mesh Adoption Patterns
- 4:01Messageoriented middleware vs Service Mesh
- 6:31Advantages of Mesh Architecture
- 8:41Who should consider a Service Mesh
- 9:28Service Mesh features
- 13:30Access logs
- 15:15Service Mesh
- 19:23Virtual Cubelet
- 21:26Ingress
- 24:20Service Mesh Complexity
- 28:56Distributed Tracing
- 33:15Context Propagation
related talks
47mService mesh options with Istio, Linkerd, and/or Consul
43mService Mesh and the future of networking
36mBringing the Envoy Service Mesh to Mobile
05mKeynote: Service Mesh Everywhere
41mPanel Discussion - The Future of Service Mesh: Is eBPF a Silver Lining or a Silver Bullet
28mEnvoy at Square
26mNavigating the Service Mesh Landscape - Paul Bouwer, Microsoft
34mIntro: Envoy