THIS IS A TEST INSTANCE. ALL YOUR CHANGES WILL BE LOST!!!!
...
To verify this optimization, we internally designed test scenarios for validation.
The agent process will continuously submit sql queries to the SQL Gateway service using different concurrency (1 concurrency, 32 concurrency and , 64 concurrency and more) and monitor the end-to-end Latency.
...
Notice: The baseline version of Flink is not exactly the same as the community version, and it includes some of our internal enhancements , like the HA improvement mentioned in FLIP-403.
E2E Latency
This benchmark shows these optimizations can reduce more than 50% E2E latency of these short-lived queries.
...
In high concurrency scenarios, the CPU usage of the Flink cluster was reduced by 20% with no change performance degradation in QPS by using proper retry strategy configuration.
Baseline(128 Concurrency) | Experiment(128 Concurrency) | |||
|---|---|---|---|---|
Configuration | DEFAULT_ RETRY_ MILLIS | 100 ms | collect-strategy.type | incremental-delay |
initial-delay | 200 ms | |||
increment | 100 ms | |||
max-delay | 700 ms | |||
Join QPS | 47 | 48 | ||
JM CPU | 17.3 |
| 14.5 |
|
TM CPU | 9.0 | 7.2 | ||
Changed Public Interfaces
...