RE: [squid-users] Multiple Accerlerating Servers

I am trying to do something similar; I haven't yet, but have an idea -- in
theory -- of how this would be done.

I think you would need to set up the accel boxes to use ICP and act as
cache-peers across a dedicated connection on its own VLAN or a crossover
connection between the two. Cache misses would first be sent as ICP queries
to the cache peer/sibling, and use that instead of the backend... I'm not
sure what you need to do to get ICP working in a accel setup, but I know it
can be done.

You would likely want to load-balance your 2 squid boxes with a L4 switch of
some kind... Some L4 switches, like Cisco and Intel have direct send return
or out of path return to have squid directly send replies back to the
originating client, rather than your load-balancer acting as a proxy of
sorts.

Sean

-----Original Message-----
From: AJ Lemke [mailto:aj.lemke@securitylabs.com]
Sent: Thursday, December 06, 2001 8:44 AM
To: Squid-Users@Squid-Cache. Org
Subject: [squid-users] Multiple Accerlerating Servers

I now have 2 squid boxes running in accel mode. I am wondering the most
efficient way to have them cache out a website without burying the backend
server with requests. Is there a way to make Squid 1 transfer its content
to Squid 2 or vice-versa, or do both Squids have to make requests to the
backend server.

I am running 2.4.stable2 and yesterdays development release of 2.5.
Received on Thu Dec 06 2001 - 10:11:10 MST