Andrew Wong has posted comments on this change. ( 
http://gerrit.cloudera.org:8080/16284 )

Change subject: [benchmarks] added a few 'perf stat' metrics
......................................................................


Patch Set 3:

(2 comments)

http://gerrit.cloudera.org:8080/#/c/16284/3//COMMIT_MSG
Commit Message:

http://gerrit.cloudera.org:8080/#/c/16284/3//COMMIT_MSG@11
PS3, Line 11:   * instructions per cycle
            :   * CPU utilization
            :   * branch misses
These are useful in getting to the bottom of why something is slow. Have you 
run these enough times to have a feel for whether they are stable enough to 
begin tracking by default? From a perf regression tracking perspective, do 
these provide more coverage than the existing latency/throughput graphs?


http://gerrit.cloudera.org:8080/#/c/16284/3/src/kudu/scripts/benchmarks.sh
File src/kudu/scripts/benchmarks.sh:

http://gerrit.cloudera.org:8080/#/c/16284/3/src/kudu/scripts/benchmarks.sh@573
PS3, Line 573:   # parse scan timings for scans and inserts with MRS only
We aren't collecting stats for a few of these tests. Was that deliberate? Where 
do we draw the line before collecting stats? What qualifies a test for it being 
useful to track its perf stats? How CPU-intensive/branch-heavy the test is?



--
To view, visit http://gerrit.cloudera.org:8080/16284
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: kudu
Gerrit-Branch: master
Gerrit-MessageType: comment
Gerrit-Change-Id: Ib742ffbaabca96c2ac455481313348da2133ff6e
Gerrit-Change-Number: 16284
Gerrit-PatchSet: 3
Gerrit-Owner: Alexey Serbin <[email protected]>
Gerrit-Reviewer: Alexey Serbin <[email protected]>
Gerrit-Reviewer: Andrew Wong <[email protected]>
Gerrit-Reviewer: Bankim Bhavsar <[email protected]>
Gerrit-Reviewer: Grant Henke <[email protected]>
Gerrit-Reviewer: Kudu Jenkins (120)
Gerrit-Comment-Date: Mon, 10 Aug 2020 20:29:11 +0000
Gerrit-HasComments: Yes

Reply via email to