[squid-users] Cache Load Balancing

Hi there,

I would just like to understand how a Sibling configuration can really be
used to distribute load on two Cache Engines.
We currently have 2 Cache Engines which pass thru different IPLs going to
the Internet. We are currently implementing WCCP to redirect all HTTP
traffic on the cache servers.

The cache engines have also been configured so that both act as a sibling
peer to the other. So this should mean if an object doesn't exist locally,
a cache engine will attempt to get it from its peers. If the cache produces
a HIT from its peers, it will not cache the object. This is to save cache
space. But if its peers produces a MISS, it will fetch the object directly
using its gateway to the Internet.

My problem is this...
I noticed that whenever I access a site and the request is redirected to a
cache server, any succeeding requests for that web site will always go
through that particular cache engine. For me this means, very low UDP hit
ratios since requests for a particular web site will always go through only
one cache engine.

I really don't know if our router redirecting the http packets to the cache
engines has something to do with this? Will BGP affect the 'cache server'
selection process? Or in other words, will the router choose the cache
engine which is closest to the web site? Or is this something to do with my
Squid's Sibling configuration?

I know you guys are very busy. But I will surely appreciate any inputs.

Thank you very much!!!

Egie
Received on Thu Aug 16 2001 - 00:16:59 MDT