How Do Big Internet Companies Achieve Cross‑Region Multi‑Active HA?

Overview

Cross‑region multi‑active (异地多活) is a high‑availability deployment model adopted by large internet companies such as Alibaba, Tencent, Baidu, NetEase, and Sina. It evolves through cold backup → dual‑machine hot standby → same‑city active‑active → cross‑city active‑active → cross‑region multi‑active.

Stateful vs Stateless Services

Stateless services achieve HA simply via load balancers (F5, etc.). Stateful services store data on disk or memory (e.g., MySQL, Redis) and require more complex solutions.

Cold Backup

Cold backup copies data files while the service is stopped. Advantages: simple, fast backup and restore, point‑in‑time recovery. Drawbacks: service downtime, data loss between backup and restore, full‑copy overhead, and inability to customize partial backups.

Dual‑Machine Hot Standby

Active/Standby mode keeps one node serving traffic while the other replicates data. Replication can be software‑based (MySQL master/slave, SQL Server transactional replication) or hardware‑based (disk mirroring). Variants include dual‑machine mutual backup, which swaps primary/secondary roles per service.

Same‑City Active‑Active

Extends hot standby within the same city, using dedicated lines for fast synchronization. It can support true active‑active reads/writes if conflict resolution is handled, but still relies on a nearby data center for disaster recovery.

Cross‑City Active‑Active (Two‑Site)

Deploys front‑end entry points and applications in two cities. When one city fails, traffic is fail‑over to the other, possibly with degraded performance due to latency.

Cross‑Region Multi‑Active Architecture

Each node has four inbound/outbound connections; any single node failure does not affect service. However, longer write latency and higher conflict probability increase complexity. Solutions include sharding by region, using distributed locks, or adopting a “Global Zone” where writes go to a master region and reads can be served locally.

Real‑World Examples

Eleme’s Global Zone, Alibaba’s three‑center model, and Taobao’s unit‑based multi‑active design illustrate different trade‑offs between consistency, throughput, and operational complexity.

Key Challenges

Increased latency for cross‑region writes.

Data conflicts requiring distributed transactions or sharding.

Higher operational overhead for testing, automation, and disaster‑recovery drills.

Discussion Questions

How would you route a user located at the intersection of four cities to the correct shard?

Which business modules in your system can be made multi‑active and which cannot?

Is multi‑active required for all services or only for core business?

References:

Eleme “Cross‑Region Multi‑Active Technical Implementation (Part 1)” – https://zhuanlan.zhihu.com/p/32009822

Eleme Architecture Blog – https://zhuanlan.zhihu.com/eleme-arch

Alibaba “Cross‑Region Multi‑Active vs Same‑City Active‑Active Architecture Evolution” – https://www.sohu.com/a/158859741_444159

Alibaba Cloud “Database Cross‑Region Multi‑Active Solution” – https://help.aliyun.com/document_detail/72721.html

“Cross‑Region Multi‑Active Is Not That Hard” – https://wely.iteye.com/blog/2313293