[squid-users] reverse proxy cache optimization

Hi,

We have 9 memory only squid proxies in front of two webserver sharing
behind a nas storage with a lot of static content to cache under very
high load.

cache_peer xxx1 parent 80 0 originserver no-netdb-exchange no-query
no-digest round-robin
cache_peer xxx2 parent 80 0 originserver no-netdb-exchange no-query
no-digest round-robin
cache_peer xxx3 sibling 80 3130 weight=1 no-netdb-exchange no-digest
proxy-only
cache_peer xxx4 sibling 80 3130 weight=2 no-netdb-exchange no-digest
proxy-only
cache_peer xxx5 sibling 80 3130 weight=3 no-netdb-exchange no-digest
proxy-only
cache_peer xxx6 sibling 80 3130 weight=4 no-netdb-exchange no-digest
proxy-only
cache_peer xxx7 sibling 80 3130 weight=5 no-netdb-exchange no-digest
proxy-only
cache_peer xxx8 sibling 80 3130 weight=6 no-netdb-exchange no-digest
proxy-only
cache_peer xxx9 sibling 80 3130 weight=7 no-netdb-exchange no-digest
proxy-only
cache_peer xxx10 sibling 80 3130 weight=8 no-netdb-exchange no-digest
proxy-only

The proxy-only parameter should already help to keep the caches more
independent from each other, if i understood it right.

What else can i do to keep the caches more unique, like i thought about
working with acl to don't cache certain directories, so that proxy a, b
and c are not allowed to cache objects from /2006/ and proxy d, e and f
the same for 2007, so they build up more unique caches because they dont
share the objects from each other.

But i don't know if i put in the no_cache option if the proxy still ask
the neighbour caches or directly go for the parents in the back?

Maybe any other ideas how i can reduce the load of the backend machines,
maybe i make a terrible mistake in the configuration and there is
another way to keep things running.

greetings

markus
Received on Tue Aug 28 2007 - 05:44:48 MDT