Design and Practice of Multi‑Active Architecture on Public Cloud Infrastructure

This article is a transcript of a talk by Wang Xiaobo, CTO of Tongcheng-Elong's ticket business group and a Tencent Cloud TVP expert, delivered at the Cloud+ Community Developer Conference (Hangzhou). The speaker shares how to simplify the design and implementation of a "multi‑active" (active‑active) architecture using public‑cloud infrastructure.

The talk begins with an introduction to the concept of multi‑active, explaining why organizations pursue it, the difficulties involved, and how cloud services can reduce the effort required. It then moves to practical considerations, starting from the physical data‑center level: power supply, building protection, network topology, and hardware layout. Real‑world observations from a German data‑center visit illustrate the importance of redundancy, power‑failure handling, and the limits of traditional backup solutions.

Key technical challenges discussed include:

Ensuring real‑time data synchronization across geographically distributed sites.

Managing network latency and avoiding single points of failure in core switches and backbone links.

Designing fault‑tolerant infrastructure with independent cooling, power, and networking modules.

Balancing capacity, cost, and resource availability when expanding to multiple sites.

The speaker emphasizes the need to move from monolithic applications to micro‑service architectures to achieve true multi‑active capabilities, while also highlighting new problems such as service discovery, circuit breaking, rate limiting, and cross‑region data consistency.

Further topics covered are disaster recovery versus capacity expansion, the trade‑offs between same‑city (low‑latency) and cross‑city deployments, and the economic considerations of building multi‑active systems versus relying on cloud‑based solutions.

In conclusion, the talk encourages attendees to carefully evaluate whether a 100% multi‑active setup is necessary, to consider the cost‑benefit balance, and to adopt incremental, cloud‑assisted strategies for improving availability and scalability.