[p4] Server performance questions.

John D. Mitchell johnm-perforce at non.net
Fri Oct 19 10:56:30 PDT 2001

>>>>> "Eric" == Eric Dew <ericd at resilience.com> writes:
> I agree with Stephen Vance's suggestion to look for the root causes.


[...Switch from beefy NT box to Sun box...]

You could also convert you x86 box to one of the xBSD or Linux OSs.

With that many simultaneous users, I'd start by looking at memory abuse.
MS NT has a pretty horribly bad memory management especially anywhere near
the limit.

Anecdote: We wrote and ran some semi-non-trivial (extracted and tweaked
from a real application) benchmarks a few years ago and the NT box ran over
the weekend and we ended up killing it Monday afternoon.  The old Linux
v2.0.x box was done in about 30 minutes.  Note that this was on the exact
same hardware -- that was a dual-boot machine.

If you're trying to actully fix the NT box rather than just switching to
something reliable, the second area that I'd look at would be the disk
configuration and setup.

ObPeeve:  I have never really understood why some folks are willing to
trust their *code* to something that is so inherently insecure and as
unreliable as an MS OS.  Boggles my mind.  :-)

Take care,

More information about the perforce-user mailing list