Understanding RPC Architecture: Basic Structure and Technical System

When building a distributed system, one must consider how services call each other; if existing frameworks like Dubbo or Spring Cloud are unavailable, implementing a custom RPC framework becomes necessary.

RPC Architecture Basic Structure

The architecture consists of providers (servers) and consumers (clients). Communication relies on a RpcChannel and RpcProtocol , while the service entry point is exposed via RpcAcceptor and requests are sent through RpcConnector . A unified contract called Remote API defines the interface.

Key components include:

Client side: RpcClient , RpcProxy , RpcCaller , RpcConnector

Server side: RpcServer , RpcInvoker , RpcProcessor , RpcAcceptor

Shared: RpcProtocol , RpcChannel

Technical System of RPC Architecture

Network Communication

Performance favors long-lived TCP connections and non‑blocking I/O (NIO) with multiplexing, while reliability requires link health checks and reconnection logic.

Serialization

Data is serialized for transmission; common choices are text‑based (XML, JSON) and binary (Protocol Buffers, Thrift). Selection should consider space, time, and CPU/memory overhead.

Transport Protocol

RPC can operate over various OSI layers; TCP serves at the transport layer, while custom protocols (e.g., Dubbo) may be built atop it to reduce overhead and increase extensibility.

Remote Invocation

Calls can be synchronous (blocking) or asynchronous using Java Future (Future‑Get or Future‑Listener). Asynchronous patterns improve throughput, while synchronous calls simplify development.

Summary

RPC is a foundational technology for service‑to‑service interaction in distributed systems; understanding its basic structure and technical stack—network communication, serialization, transport protocols, and invocation models—helps developers design and implement effective remote call mechanisms.