Case Study: Mogujie achieves 600,000 requests/sec and low-latency for 60M shoppers with ScyllaDB

Building Data Store to Support 60 Million Online Shoppers

Mogujie is a major fashion retail site serving over 60 million online users and more than ¥3 billion in turnover, and needed a resilient, low‑latency data store able to sustain roughly 50,000 TPS with smooth scaling, high availability, and no single point of failure for mixed read/write workloads. After problems with Apache Cassandra and HBase, Mogujie evaluated and adopted Scylla (from ScyllaDB) to meet those requirements.

ScyllaDB’s Scylla was deployed on a six‑node cluster and immediately exceeded expectations, delivering 100,000 TPS in tests and later serving up to 600,000 requests per second in production. Following ScyllaDB’s guidance Mogujie moved to SSDs and optimized their schema (inserting 360 data points at a time) to reach about 100 million data points per minute, shift bottlenecks away from disk to the network, support 10 TB per node, reduce operational and capital costs, and run ScyllaDB in production for more than a year with low latency and high availability.

Mogujie

FengLin (MengYe Shen)

Mogujie