jmeter-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From GitBox <>
Subject [GitHub] [jmeter] pmouawad commented on pull request #603: Bug 64558 - Improve performances and throughput of Sample Results by lifting contention on writing SampleResults in CSV/XML
Date Wed, 08 Jul 2020 20:30:27 GMT

pmouawad commented on pull request #603:

   > > > Where you previously had all threads waiting on lock to write to CSV , it
is not more present.
   > > > This is particularly the case in a plan with no Think Time at low number
of threads, or a plan with think times but at high >number of threads (5000 in my case).
   > > 
   > > 
   > > Do you have the results for the throughput / sustained operation rate?
   > > I agree there might be improvements, however, I'm afraid, the improvements from
the PR like this can't exceed 2-5%, and it is strange to hear "performance increased by 68%".
   > > Of course, there might be edge cases when the improvement is 68%. However, it
does not seem to be right to declare a major performance improvement for an unrealistic case.
   > In the approach I followed, I am tuning a piece of code, to see improvements on it,
I need to build a case where it is the only (if possible) piece of code involved, and based
on this I conclude whether optimization improves or not.
   > If I add other factors that could slowdown (HTTP) , how can I evaluate whether good/bad
results come from my code of from all involved factors when using HTTP.
   > Second , in the test I proposed, you can adjust the waiting time of the Java Request
to simulate a real life call.
   > I also made test with 10ms waiting time when working on it. Do you want me to share
those results ?
   > Why don't you try it in your configuration and see what you get ?
   > If it's just a communication matter and you don't want to mention 68% improvement,
no problem for me, I don't want to loose too much time on this.
   > If it's another matter then please clarify.
   > Thanks
   It looks like I mixed up results previously.
   With 10ms we have 68% improvement, while there is no improvements but degradation with
0ms sleepTime.
   Anyway, I guess it would be better that reviewers of this PR make their own tests to double

This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:

View raw message