[squid-users] Re: performance

azeem ahmad wrote:

> since the day i got my cache full (two partitions of 3 gb each) i m facing
> the problem of low speed.

If the proxy is slow after the cache fills up, then your system is likely
running out of RAM and beginning to swap. What is the output of free?

> is it the problem of

> 1- cache replacment policy (i m using squid2.5stable1-2 came with Red Hat
> 9 distribution. should i recompile it to use heap LFUDA. i m using heap
> LFUDA as cache and memory replacement policies but i havent recopiled)
> 2- heavy squid logs (swap.state and access log etc)

If the RPM has the features you want, then there's no need to recompile.
Neither of these is likely the cause of your problem.

> how can i tune it up and how i can check the hit ratio.

Tune for system performance or tune for caching performance?

Best system performance will be had by making sure your system does not run
out of memory and start to swap. See the Memory FAQ for more information:

http://www.squid-cache.org/Doc/FAQ/FAQ-8.html

Also, if possible you should keep your cache on its own dedicated disk. The
best file system for the cache is ReiserFS.

Caching performance can be improved by adjusting the refresh_pattern
settings in squid.conf. See the squid.conf.default file for details on how
these settings work.

Calamaris (linked to from the Log Analysis page on squid-cache.org) can give
statistics on hit ratios by both bytes and requests.

> an any one tell me the difference of tcp_hit, tcp_refresh_hit, tcp_ims_hit

See the Squid result codes FAQ:

http://www.squid-cache.org/Doc/FAQ/FAQ-6.html#ss6.7

Adam
Received on Sat Oct 02 2004 - 18:28:43 MDT