Panel Discussion: Ask Us Anything: Microservices and Service Mesh
summary
Panel covers why service mesh exists and concrete trade-offs: solves routing, retries, mTLS and observability but adds CPU/latency cost; examples include an API product using 170 microservices to gain telemetry and automated encryption, an ad firm running ~30 services for latency and resiliency, and production Linkerd use on GKE for high-value traffic. Practical notes: sidecars can add about 8ms extra hop in some profiles, gateways/VM integrations bridge legacy workloads, and tools like Envoy, gRPC, WASM and Flagger/CoopCost sit on top to manage rollout, cost, and performance.
AI-written summary · about
chapters (13)
- 0:00Introduction
- 3:41Why Service Mesh
- 4:47Microservice Architectures
- 5:24Security and Reliability
- 7:45Service Mesh Production
- 10:45LinkerD Users
- 12:01Service Mesh Use Cases
- 13:42Service Mesh Path Latency
- 16:03Question from the back
- 18:55Sidecar performance
- 23:27Patterns and tips
- 27:16Perspective
- 33:31CPU footprint
related talks
36mPanel: Ask Me Anything: Microservices and Service Mesh [B]
38mAsk Me Anything About Service Mesh
45mService Mesh War Stories - Lessons Learned the Hard Way
43mWTF Do We Need a Service Mesh?
40mComparison of Current Service Mesh Architectures
46mWhat the hell is service mesh anyway !?
35mNavigating the Service Mesh Ecosystem
56mService Meshes - Powering the next wave of microservice architectures