How WeChat Scales: Backend Architecture for Ultra‑Low Latency Messaging

Note: Company lecture notes, not guaranteed accurate.

Problem

Extreme business features

Smooth message sending and receiving

Timely notifications

Power saving

Bandwidth saving

Thin client

Difficult backend‑terminal synchronization

Synchronizing diverse data: account info, contacts, messages, moments, etc.

Timely notification and sync

Reliable sync over mobile networks

Saving bandwidth and power

Solution

Minimalist sync protocol

The backend and terminal only need to exchange a single number; the backend can infer all missing data on the terminal.

Change sequence number / version number:

Efficient notification mechanism

iOS Apple Push Network Service

Android and others – long connections

GPRS/EDGE signaling storm optimization

Adaptive heartbeat interval adjustment

Three‑layer backend architecture

Unified RPC framework

Generate server and client code from Protocol Buffer definitions

Hide network details

Rich features

High‑concurrency coroutine RPC

Server‑side synchronous call model is easier to learn, use, and debug than an asynchronous model, but a single server can only support a limited number of processes and threads.

RPC based on user‑level threads (coroutines)

A single machine can support tens of thousands to a hundred thousand user‑level threads, limited only by CPU and memory.

Improves concurrency and performance.

Implementation of user‑level thread RPC

Based on makecontext / getcontext / swapcontext Hook network calls: read/write/epoll

User‑level thread scheduling

Proximity access

IDC proximity access

Network access near major carriers

CDN for image upload/download

Multiple IDC distribution improves user experience

Complex domestic network environment

Over 100 million overseas users

Each IDC provides full functionality and all required data

Each IDC holds both shared and independent data

IDC distributed data high‑reliability eventual consistency guarantee

Primary‑backup model for account and SNS data

Weakly real‑time cross‑IDC updates use a Zookeeper‑mediated primary‑backup task queue

Social graph cross‑IDC updates

Fault tolerance and disaster recovery mechanisms

Single IDC

High‑availability geographic disaster recovery

Source: http://blog.xiayf.cn/2013/10/23/learning-in-tencent-backend-arch-of-weixin/