Steve James:
We've experienced a slow but steady degradation of performance (based on
throughput, average response time for p4 sync, etc) over the last year
with
our existing hardware (dual PIII 800MHz, 4GB ram, RAID1 UltraSCSI disk,
NT4
Enterprise /3GB option). We have about 350 users (150-200 of which are
very
active), our db.* files are roughly 14gb total.
We recently upgraded to 2002.1 and implemented MaxResults and
MaxScanRows,
and most of the other best practices. Steady state performance of the
server is definitely better than with 2001.1, but under heavy load
2002.1
actually seems worse to us, we've had 4 episodes of the server just
grinding
to a halt in the past 2 weeks.
Have you checked the actual performance of the server (via perfmon or
something)? My guesses for where to look for improved performance
would be:
1. Disk optimization (even a fast disk wastes a lot of time if it has
to look all over creation for a sector). You might want to see about
dividing up the load, too - give windows its own drive for temp and
swap files, for example. (And remember that partitions are not real
drives :-)
2. Network throughput
3. RAM bus speed (I used to write graphics routines and I could max out
the bus throughput long before I broke a sweat optimizing the CPU
usage.)
-Wm
4 messages in com.perforce.perforce-user[p4] performance gains with beefier h...