Case Study: JD Cloud achieves 40K QPS and 20+ Billion Rows in a Single Cluster with PingCAP TiKV

How We Achieved 40 K QPS and 20+ Billion Rows of Data in a Single Scale-out Cluster

JD Cloud & AI, the cloud computing and AI arm of JD Group, faced significant challenges in managing the explosive growth of its Object Storage Service (OSS) metadata. Their initial MySQL-based system was unable to horizontally scale, complicating data management as they anticipated needing to store hundreds of billions, or even a trillion, rows of data. This limitation prompted their search for a new, scalable database solution, leading them to PingCAP.

PingCAP's distributed key-value database, TiKV, provided the ideal solution with its globally ordered key-value store and horizontal scalability. JD Cloud & AI successfully migrated their metadata to TiKV, achieving a peak of 40,000 QPS with 10-millisecond latency in a single production cluster storing over 20 billion rows of data. The implementation simplified their storage architecture and provided the robust, scalable foundation needed for future data growth, enabling them to plan for MySQL's retirement.

JD Cloud & AI

Can Cui

Infrastructure Specialist