gRPC — Part 2: Architecture, HTTP/2, Protobuf & Streaming

What is this?

In Part 1 we said gRPC lets you call a remote method like a local function — the phone extension to a colleague's desk. This part is about how the phone line is built. Why does the call feel instant? Why is the payload so small? How can both people talk at once without taking turns?

The one-line idea: gRPC's speed comes from two engineering choices stacked together — HTTP/2 (one connection, many simultaneous conversations) underneath Protocol Buffers (compact binary instead of chatty text) — and its flexibility comes from four streaming modes layered on top.

Let's open each layer.