Case Study: Yum China achieves low-latency, high-throughput messaging with StreamNative Pulsar

Pulsar at Yum China Achieving Low Latency and High Throughput for Business and Operational Data

Yum China, one of China’s largest restaurant companies, needed a message queue platform that could scale horizontally while keeping operational costs low. The company used StreamNative’s Apache Pulsar to support internal business and operational data workloads with very different requirements for durability, latency, and throughput.

Using StreamNative’s Pulsar-based setup, Yum China built separate clusters for business and ops traffic, then tuned them for their specific needs with synchronous sends, replication, journal sync, and low-latency storage for core data, and asynchronous sends for monitoring and other ephemeral workloads. StreamNative’s tooling and Pulsar architecture helped Yum China cut maximum write latency from 300 ms to 5 ms by moving to local SSDs, reduce GC pauses to around 10 ms, lower maximum publish latency from 500 ms to 100 ms, and reach up to 200,000 messages per second for monitoring ingestion and 46,000 messages per second per node in optimized throughput scenarios.

Yum China

Chauncey Yan

Backend Software Engineer