graphite-dev team mailing list archive

[chrismd <question170521@xxxxxxxxxxxxxxxxxxxxx>]

Question #170521 on Graphite changed:
https://answers.launchpad.net/graphite/+question/170521

    Status: Open => Answered

chrismd proposed the following answer:
I don't have a canned test script for you, I usually put together a
custom test each time for whatever it is I'm testing but I've not
assembled a test suite of any sort.

Are you just trying to figure out your throughput ceiling for your
hardware? If so you should definitely do a full end-to-end test
including carbon as well as graphing requests against the webapp. That
is the best way to get a realistic answer.

Or are you just curious how whisper performs as a piece of software? Honestly I don't think there is much point in load testing whisper this way because:
1) I/O performance vastly overshadows CPU/memory usage
2) whisper is not optimized for CPU/memory usage, if for no other reason than its python. It scales simply because of the I/O & buffering strategy (which is managed by carbon, not whisper).
3) whisper's I/O patterns are very predictable and so doing a whisper load test basically becomes a test of how well your system's disks perform with that type of I/O load. If this is what you want again I would say do a full end-to-end test including carbon to get realistic results.

I hope that helps.

-- 
You received this question notification because you are a member of
graphite-dev, which is an answer contact for Graphite.