gRPC Deep Dive
summary
Explains gRPC’s C-core as three layers (surface with API and filter stack, transport, IO manager) and traces a unary RPC: channel → multiple subchannels (one per backend) → call creation → batched operations posted to a completion queue → server-side batch handling and response. Covers client-side load‑balancing policies (pick_first, round‑robin, gRPC LB), lazy connection behavior, retries and service‑config as ongoing work, measurable faster C/C++ impl versus Java/Go, and multithreaded completion‑queue threading for performance.
AI-written summary · about
chapters (8)
- 0:00Start
- 0:24Target Audience For This Talk
- 3:16GRPC C-Core: Architecture Overview
- 4:47GRPC C-Core Programming model
- 7:02Quick recap: Ops and Messages between Client and Server
- 8:41Life of an RPC: Channel Creation
- 15:26Life of an RPC: Overview
- 20:49Conclusion
