[
https://issues.apache.org/jira/browse/FLINK-19441?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17214410#comment-17214410
]
Yingjie Cao commented on FLINK-19441:
-------------------------------------
As a first step, I tried to run the two benchmark cases on commit c3fab5173c
([FLINK-19391][network] Moved notification during subpartition request to the
requester. [Arvid Heise]) and commit aa62e64902 ([FLINK-19323][network] Small
optimization of RecordWriter#serializeRecord [kevin.cyj]). There is no
regression and the performance is even better for tupleKeyBy. For globalWindow,
there is no visible different. The results are as follows:
To avoid influence from other processes, I closes all other applications on my
laptop, after which only about 3% CPU is used. Besides, two versions of Flink
package ran in turn, the odd line in the following results is from commit
c3fab5173c and the even line is from commit aa62e64902 (line number starts from
1).
```
(base) kevin.cyj@kevincyjs-macbook-pro flink-benchmarks % grep
"KeyByBenchmarks.tupleKeyBy" ../result.txt | grep "thrpt"
KeyByBenchmarks.tupleKeyBy thrpt 30 4399.725 ± 180.287 ops/ms
KeyByBenchmarks.tupleKeyBy thrpt 30 4393.090 ± 215.023 ops/ms
KeyByBenchmarks.tupleKeyBy thrpt 30 4646.292 ± 273.936 ops/ms
KeyByBenchmarks.tupleKeyBy thrpt 30 4123.914 ± 108.549 ops/ms
KeyByBenchmarks.tupleKeyBy thrpt 30 4349.349 ± 275.403 ops/ms
KeyByBenchmarks.tupleKeyBy thrpt 30 4808.382 ± 362.876 ops/ms
KeyByBenchmarks.tupleKeyBy thrpt 30 4224.644 ± 236.571 ops/ms
KeyByBenchmarks.tupleKeyBy thrpt 30 4649.511 ± 255.074 ops/ms
KeyByBenchmarks.tupleKeyBy thrpt 30 4340.121 ± 230.217 ops/ms
KeyByBenchmarks.tupleKeyBy thrpt 30 4973.276 ± 161.522 ops/ms
KeyByBenchmarks.tupleKeyBy thrpt 30 4481.989 ± 250.952 ops/ms
KeyByBenchmarks.tupleKeyBy thrpt 30 4494.742 ± 328.143 ops/ms
KeyByBenchmarks.tupleKeyBy thrpt 30 4482.170 ± 245.795 ops/ms
KeyByBenchmarks.tupleKeyBy thrpt 30 4210.755 ± 136.354 ops/ms
KeyByBenchmarks.tupleKeyBy thrpt 30 4244.761 ± 255.086 ops/ms
KeyByBenchmarks.tupleKeyBy thrpt 30 4687.526 ± 299.669 ops/ms
KeyByBenchmarks.tupleKeyBy thrpt 30 4188.102 ± 240.958 ops/ms
KeyByBenchmarks.tupleKeyBy thrpt 30 4658.756 ± 257.802 ops/ms
KeyByBenchmarks.tupleKeyBy thrpt 30 4370.287 ± 206.037 ops/ms
KeyByBenchmarks.tupleKeyBy thrpt 30 4583.454 ± 286.767 ops/ms
(base) kevin.cyj@kevincyjs-macbook-pro flink-benchmarks % grep
"WindowBenchmarks.globalWindow" ../result.txt | grep "thrpt"
WindowBenchmarks.globalWindow thrpt 30 5359.334 ± 92.793 ops/ms
WindowBenchmarks.globalWindow thrpt 30 5473.884 ± 72.067 ops/ms
WindowBenchmarks.globalWindow thrpt 30 5449.328 ± 88.903 ops/ms
WindowBenchmarks.globalWindow thrpt 30 5316.775 ± 307.340 ops/ms
WindowBenchmarks.globalWindow thrpt 30 5413.462 ± 89.206 ops/ms
WindowBenchmarks.globalWindow thrpt 30 5403.000 ± 337.460 ops/ms
WindowBenchmarks.globalWindow thrpt 30 5409.112 ± 92.309 ops/ms
WindowBenchmarks.globalWindow thrpt 30 5556.913 ± 103.412 ops/ms
WindowBenchmarks.globalWindow thrpt 30 5512.030 ± 95.002 ops/ms
WindowBenchmarks.globalWindow thrpt 30 5490.936 ± 87.189 ops/ms
WindowBenchmarks.globalWindow thrpt 30 5520.000 ± 83.022 ops/ms
WindowBenchmarks.globalWindow thrpt 30 5422.663 ± 107.169 ops/ms
WindowBenchmarks.globalWindow thrpt 30 5353.005 ± 91.061 ops/ms
WindowBenchmarks.globalWindow thrpt 30 5451.412 ± 104.578 ops/ms
WindowBenchmarks.globalWindow thrpt 30 5444.123 ± 89.937 ops/ms
WindowBenchmarks.globalWindow thrpt 30 5512.661 ± 64.927 ops/ms
WindowBenchmarks.globalWindow thrpt 30 5377.912 ± 89.680 ops/ms
WindowBenchmarks.globalWindow thrpt 30 5480.517 ± 68.624 ops/ms
WindowBenchmarks.globalWindow thrpt 30 5373.978 ± 122.724 ops/ms
WindowBenchmarks.globalWindow thrpt 30 5436.810 ± 102.140 ops/ms
```
> Performance regression on 24.09.2020
> ------------------------------------
>
> Key: FLINK-19441
> URL: https://issues.apache.org/jira/browse/FLINK-19441
> Project: Flink
> Issue Type: Bug
> Reporter: Arvid Heise
> Assignee: Stephan Ewen
> Priority: Blocker
> Labels: pull-request-available
>
> A couple of benchmarks are showing a small performance regression on
> 24.09.2020:
> http://codespeed.dak8s.net:8000/timeline/?ben=globalWindow&env=2
> http://codespeed.dak8s.net:8000/timeline/?ben=tupleKeyBy&env=2 (?)
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
