Inside Meitu’s Backend: How 1.5 B Users Power Meipai’s High‑Concurrency Systems

On January 20, 2018, Meitu held the 8th Internet Technology Salon in Shenzhen, live‑streamed on Meipai, featuring senior engineers who dissected the backend technologies that enable Meipai’s massive user base.

High‑Concurrency Prop‑Trading System

Guo Chenhui, Technical Director of Meipai, explained the architecture behind the prop‑trading platform that handles millions of simultaneous orders. He highlighted challenges such as transaction bottlenecks, data consistency, and latency, and described optimization techniques like sharding, asynchronous processing, and cache strategies that keep the system stable under heavy load.

Comprehensive Monitoring Platform

Wang Guansheng, Technical Director of Operations, presented Meitu’s multi‑layered monitoring framework. He emphasized the importance of proactive alerts, detailed post‑incident data, and the integration of AIOps for automated anomaly detection. The talk covered a unified alarm platform based on business‑level dependency graphs, standardized reporting, and large‑screen dashboards, as well as ongoing work on an abnormal‑detection platform.

IM System Evolution

Lai Jianbing, Senior Technical Manager, traced the Meitu IM system from its inception to a generalized architecture. He discussed early design decisions, performance‑tuning measures for high availability, and the trade‑offs made during the transition to a more reusable, modular framework, concluding with future plans for new features and further performance gains.

Live‑Streaming Architecture Optimization

Mai Junsheng, Technical Director of the Architecture Platform, examined the challenges of live‑streaming at scale. He described the initial technology selection, subsequent bottlenecks, and the self‑developed optimizations across client SDKs, CDN and origin server monitoring, and network probing. By aggregating telemetry data, Meitu built a full‑stack live‑stream monitoring system and began evolving toward multi‑CDN intelligent scheduling to improve latency and reliability.

Conclusion

The salon demonstrated Meitu’s commitment to sharing cutting‑edge backend practices, from high‑throughput transaction processing to AI‑enhanced operations monitoring, offering valuable insights for engineers building large‑scale, high‑availability services.