gRPC & Kubernetes
summary
Built a hosted Prometheus monitoring service that uses HTTP/2 with protobuf gRPC for all intra-server RPCs on Kubernetes; code generation produces trivial Go stubs and a working greeter example. Replacing kube-proxy routing with kube-resolver yields true client-side load balancing for persistent connections. Front-end HTTP-over-gRPC reduces 50th‑percentile latency by ~30 ms. Swapping proto reflection with gogo-proto cuts 99th‑percentile latency from ~100 ms to ~25 ms (≈4x). Middleware adds consistent Prometheus metrics and OpenTracing across services.
AI-written summary · about
chapters (7)
- 0:00Introduction and background
- 2:10What is gRPC
- 9:15History of RPC frameworks
- 13:15gRPC basics and demo
- 17:33gRPC in Kubernetes
- 23:47Middleware and tracing
- 25:31Advanced topics and Q&A
related talks
46mIntroduction to gRPC
35mgRPC Load Balancing on Kubernetes - Jan Tattermusch, Google
40mgRPC, let's communicate differently
01h 00mScalable Microservices with gRPC & Kubernetes
21mReal World Migration from HTTP to gRPC in Ruby
28mgRPC: 5 Years Later, Is It Still Worth It?
06mKeynote: gRPC Project Update
14mOverview of gRPC