I want to know if Risingwave can achieve less than 20ms latency and 50,000 QPS for model serving with Redis sink. Currently, we use Flink sink to Redis for model serving. Can Risingwave satisfy similar scenarios?
zuston
Asked on Nov 22, 2023
Risingwave is designed for scenarios requiring low latency and high QPS. It has a row-based storage and supports creating indexes to accelerate queries. The latency target of less than 20ms can be achieved, as shown by the 99th
percentile latency metric. The QPS of 50,000 will depend on the hardware configuration of the instance used. Risingwave is continuously improving performance, and the new standalone deployment mode eliminates network communication overhead, leading to even better latency numbers in the future.