capacity planning in accelerator mode

Hello squid gurus and also to the newcomers like myself!
I'm looking for more information about configuration of our http
accelerators. Our 3 squid boxes are:
500mhz PIII
512mb RAM
9gb drive (some SCSI, some IDE)
OS(kernel)= RH linux 6.0(2.2.5-15) or 6.1(2.2.12-20)
Squid v2.3.STABLE2

Also, our cache hit according to calamari is 88.90% - does that seem high?
The cache_dir and cache_mem settings are:
cache_dir ufs /usr/local/squid/cache 400 16 256
cache_mem 128MB
Does it make sense to max out the cache_mem to ~480mb and/or bump the folder
size higher?

I'm wondering if our topology is a good choice.
We have three core (apache) servers at www1 -www3 which are round-robin
resolved via mmm.altrec.com
We then have three squid boxes www4-www6 which all use http_accel_host
mmm.altrec.com and are round-robin resolved via www.altrec.com We are
planning on adding 3 addl servers, and I was thinking that adding 2 squid
servers for each core webserver is the best way to scale, but with such a
high cache-hit maybe it makes more sense to do 3 for 1 or even more?

Thanks for any insights!

mdf += "<http://altrec.com/>";
Received on Thu May 11 2000 - 08:06:45 MDT