Hi,
I have to run performance tests on a new squid-based caching appliance.
I've done all the initial configuration, but could do with an idea if
I'm wasting my time or not...
The only machines at my disposal for testing are:
1 x celeron @ 900MHz / 256MB RAM (srv01)
1 x K6-2 @ 450MHz / 256MB RAM (clt01)
1 x P2 @ 450MHz / 64MB RAM (monitor & DNS)
(using an SMC 10/100 switch, none of the machines have fantastic NICs)
that's it... can't borrow any more for testing. Only 1 server & 1 client.
In the netperf tests I got ~50Mb/s in the bidirectional test and ~85Mb/s
(srv01 end), ~62Mb/s (clt01 end). These results are significantly below
the "should get" values listed in the docs of 80Mb/s (bi-di), 92-95Mb/s
(uni).
Our cache appliance has 2GB RAM and 80GB or 160GB of IDE disk. I imagine
it'll take a ludicrously long time for these machines to fill this
cache, which is why I seek advice now...
Will I be wasting my time to continue? If so, can anyone suggest a way
that I can test a cache like this with my limited resources?
Any help will be most gratefully received, i've been reading as much as
I can but can't find any definitive answers.
Thanks in advance,
Matt
ps. I'm running the polygraph recommended FreeBSD distro on all machines.
This archive was generated by hypermail 2b29 : Mon Feb 06 2006 - 12:00:24 MST