[
https://issues.apache.org/jira/browse/SOLR-16530?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17631228#comment-17631228
]
Ishan Chattopadhyaya commented on SOLR-16530:
---------------------------------------------
bq. e.g. What are the red and blue lines on the graph you attached a screenshot
of?
These refer to the two "tasks" in the suite
(https://github.com/fullstorydev/solr-bench/blob/ishan/repeatable-jenkins/suites/cluster-test.json).
task1 creates 1000 collections, task2 restarts 7 out of the 8 nodes in batches
of 2 (while waiting for each node to come up with all replicas active before
proceeding to the next batch. The two lines in the graph represent total time
in seconds for completing those two tasks.
bq. What's the relationship between the "screenshotted" graph and the one at
"http://mostly.cool/cluster-test-with-patch.html";?
Should be the same. I screenshotted it for posterity, because that "patch"
build and graphs are a temporary thing.
bq. (I'm assuming that the screenshot is just a static capture of the HTML
page, that you took because the dates covered by the HTML page will shift as
time passes, but an explicit confirmation there would be great.)
True.
bq. What is the "patch" referenced in the url "cluster-test-with-patch.html"?
It is the patch for removing parallelStream (which restores the performance
caused by the SOLR-16414's Nodedown NPE commit. Here's the patch:
https://github.com/fullstorydev/solr-bench/blob/ishan/repeatable-jenkins/stress.sh#L46.
Had to retroactively apply this patch and plot the graphs so that the effect
of this perf regression is nullified and other culprits (like this issue) can
be identified.
bq. What does this perf test do in a sentence or two? etc.
Goal is to measure any slowdown in collection creation performance, as well as
startup/shutdown performance in a cluster with lots of collections.
bq. Is the Y-axis on this graph in seconds?
Yes
bq. How significant is this slowdown in amortized terms?
One can deduce from the graph, depending on how one calculates. My reading is
that the JAX RS issue caused a jump from 320sec to 375sec, and this issue
caused a jump from 375sec to 475sec (assuming all other commits did not
contribute to any slowdown).
> Performance degradation due to eliminating noggit Writeable
> -----------------------------------------------------------
>
> Key: SOLR-16530
> URL: https://issues.apache.org/jira/browse/SOLR-16530
> Project: Solr
> Issue Type: Bug
> Security Level: Public(Default Security Level. Issues are Public)
> Reporter: Ishan Chattopadhyaya
> Assignee: Noble Paul
> Priority: Blocker
> Fix For: 9.2
>
> Attachments: Screenshot from 2022-11-09 11-20-35.png
>
> Time Spent: 20m
> Remaining Estimate: 0h
>
> During performance benchmarking on branch_9x, I observed a slowdown in
> restart performance since commits in SOLR-16512.
> CC [~noblepaul].
> http://mostly.cool/cluster-test-with-patch.html
> The benchmark is here:
> https://github.com/fullstorydev/solr-bench/blob/ishan/repeatable-jenkins/suites/cluster-test.json
> Effort to automate these benchmarks is WIP and tracked here: SOLR-16525.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
