Per-call latency
Sequential echo calls, 2000 per cell. p50 solid, p99 faded. Lower is better.
Throughput
8 concurrent clients × 5 seconds. Higher is better.
gRPC vs REST echo — measured three times: Rust (tonic vs axum), Go (grpc-go vs net/http), Python (grpcio vs stdlib). Payloads from 1KB to 16MB. Every number below was measured on the same 2-core VM. No simulations, no estimates.
Sequential echo calls, 2000 per cell. p50 solid, p99 faded. Lower is better.
8 concurrent clients × 5 seconds. Higher is better.