Why it matters

gRPC streaming is often the right answer for backend-to-backend real-time or batch. It combines HTTP/2 efficiency with strong typing via protobuf.

Advertisement

The architecture

Unary: one request, one response. Standard RPC.

Server streaming: one request, N responses. Server sends until done.

Client streaming: N requests, one response. Client sends until done.

Bidirectional: N in, M out, interleaved. Both stream freely.

gRPC modesUnary1 req → 1 respServer / Client stream1 → N or N → 1Bidi streamN → M interleavedAll modes use single HTTP/2 stream; protobuf gives strong typing on messages
Four gRPC modes.
Advertisement

How it works end to end

HTTP/2 multiplexing: many streams over one connection. Efficient for high concurrency.

Protobuf: strongly-typed messages. Cross-language.

Deadlines: propagate through call chains. Enable coordinated timeout.

Interceptors: middleware for auth, logging, retries.