Open owlsome2501 opened 2 months ago
the regression is the accumulation result of multiple changes
i test it again, the overall qps degradation compared with version 8.3.0 is 0.92%, and #55634 contributed 0.36% | "bench_type" | "bench_sub_type" | "tidb_version" | "thread" | "duration(s)" | "tps" | "qps" | "avg_latency(ms)" | "P99_latency(ms)" | "P999_latency(ms)" |
---|---|---|---|---|---|---|---|---|---|---|
taobench | workload_a | 8.4.0-4c23e (#55634) | 15 | 2206 | 8922 | 13637 | 13.1 | 138 | 1159 | |
taobench | workload_a | 8.4.0-d1a96(prev commit of #55634) | 15 | 2207 | 8953 | 13687 | 13.1 | 138 | 1155 | |
taobench | workload_a | 8.3.0 | 15 | 2207 | 9005 | 13764 | 13.0 | 136 | 1143 |
compared with previous commit, https://github.com/pingcap/tidb/pull/55634 introduces 2 memory allocation, and cost of read from map, this mount of degradation is expected
Bug Report
1. Minimal reproduce step (Required)
2. What did you expect to see? (Required)
No performance regression.
3. What did you see instead (Required)
QPS regression 5.98% in benchmark taobench (workload_a) compared to baseline 24060.14.
4. What is your TiDB version? (Required)
pingcap/tidb@4c23efb10b32318de7dbec7c9bec80b43b396c78