gruuya commented on issue #5504:
URL: 
https://github.com/apache/arrow-datafusion/issues/5504#issuecomment-1973862429

   Ok, I think I'm up to speed with the history of this issue now. 
   
   To me it seems that there are 2 separate action items discussed here:
   1. CI benchmark that will comment on PRs once merged to main, and possibly 
on open PRs after invoking some hook (points 1 and 2 in the issue description). 
This will be just a summary of the perf comparison between the base commit and 
the current/merge commit. This might be done using github shared runners, as 
presumably the variance wouldn't be too wild during the same run.
   2. As an extension of 1. above, a way to persist this info, and ideally 
visualize it over time. This would involve a bit more effort, as it would 
require setting up (and maintaining) dedicated infra to control for the 
variance. I guess the idea was to be checking in the [benchmark 
numbers](https://github.com/apache/arrow-datafusion/issues/6107) into 
some(this?) github repo, as opposed to a cloud DB somewhere?
   
   Also looks like both 
[conbench](https://github.com/apache/arrow/pull/39592#issuecomment-1969889622) 
and 
[bencher](https://github.com/bencherdev/bencher?tab=readme-ov-file#example-pr-comment)
 are eligible to handle the above.
   
   IMO it's best to try and tackle 1. for now, as this would present a big win, 
and would pave the way to 2 as well.


-- 
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.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to