Re: [squid-users] new squid setup, optimizing cache

From: Amos Jeffries <squid3@dont-contact.us>
Date: Thu, 29 Nov 2007 11:48:18 +1300 (NZDT)

> Hello,
> I'm setting up squid on a new machine, this one an OpenBSD 4.2 router.
> The install went fine and all, but now i come to configuring squid. I gave
> the squid cache it's own partition in this case /dev/wd0h, on the first
> drive, it's got soft updates enabled on it, and is 17.6 gb in size. I was
> wondering the best squid optimizations?
> I read that the cache size shouldn't exceed 80% of the total space so
> would that be 80% of the 17.6 g partition And when entering this on the
> cache_dir line should the size be approx 15000 or should i append
> something,
> 15g for instance?or the entire disk? For the memory and cache replacement
> policies Heap LFUDa, or is there a better choice?

the cache_dir option is still a little messy, there are no units on the
sizes there. So your initial '15000' is correct to create a 15GB
cache-dir.

Optimisations would be to use aufs to grab asynchronous capabilities of
store. And possibly a small COSS cache-dir to speed up small object
caching.

> Unrelated to cache, but my logformat i've got the squid logformat
> uncommented, i'd like to display it's time formats in local time for
> interpreting at a glance. Is this doable?

Yes. Easiest way is to "emulate_httpd on". that creates access.log in
common apache format.
Otherwise you can make your own log format.

This should help you out:
http://www.squid-cache.org/Versions/v2/2.6/cfgman/

Amos
Received on Wed Nov 28 2007 - 15:48:26 MST

This archive was generated by hypermail pre-2.1.9 : Sat Dec 01 2007 - 12:00:03 MST