graphite-dev team mailing list archive

[Steve Thompson <question166527@xxxxxxxxxxxxxxxxxxxxx>]

Question #166527 on Graphite changed:
https://answers.launchpad.net/graphite/+question/166527

    Status: Answered => Open

Steve Thompson is still having a problem:
Disks - The disks in the quads and the dodecs are server grade SSDs -
they're mirrored though, I haven't played with striping/5/10, but
they're not much of a bottleneck.

Datasets - I work for a 1600-transacations-per-minute-at-peak business.
Our backend is lots of Java and Oracle with some other bits like Redis
and MySQL thrown in. Currently we're in the process of experimenting
which is why I asked this question. I understand that hundreds of
dataset with thousands of datapoints and various retentions is going to
have a performance hit BUT my current observations are based purely on
my made up files that I mentioned in my OP. The quad core machine is
currently receiving metrics from an Oracle database collected every 30
seconds. Because of the extra overhead this would put on a production
database I don't see it being any faster. I can think of a few places
where we might go down to 10 seconds but I don't see it being more than
that.

I'm happy to provide anyone with as many details as needed but just want
to clarify that my question is "How do I make sure I'm going as fast a
possible" not "How can I go faster". If what I'm getting is top speed
then I'm happy with that - I just want to be sure. An example, although
made up and slightly silly, of what I mean is "should I rebuild Python
with ../configure --with-use-multiple-core-processors --with-go-really-
fast"?

Steve

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