This topic describes the test metrics used for the test.

The performance test uses the following four metrics:

  • Runner: the number of runners
  • Thread: the number of YCSB threads created by each runner
  • QPS (row): number of rows written or read per second
  • Avg Latency (ms): average latency calculated by runners, in milliseconds

All stress tests use synchronous requests. Therefore, the request queries-per-second (QPS) is related to the number of runners and average latency. The following content describes their relationship:

Total QPS = 1,000/ Average latency × 5 (number of runners) × Number of threads of each runner

For example, in a scenario with four partitions, five runners, five threads per runner, and an average latency of 4 ms, the theoretical QPS for a single thread is 250 (1000/4). The total QPS in this scenario is 6,250, which is calculated based on the formula: Total QPS = 250 × 5 (number of runners) × 5 (number of threads per runner).

Note This performance test is not a limit test of the service performance. The test does not trigger throttling measures on the Tablestore server. The automatic load balancing feature of Tablestore guarantees horizontal scale-up of the service capabilities provided by a single table.