ICP/CARP architecture

I am a little bit confused about using ICP/CARP with SQUID.

In my configuration, I have currently one proxy running squid, connected to
the internet. Behind this proxy, I have several branch office connected
using serial lines, some with a local proxy (NS, MS, ...) and the other
ones using my proxy directly. I have 2 questions:

1. beside my proxy, I have a second machine (currently hot standby) and I
was wondering if I could use it to load balance the load (and the space)
used by the cache. I have seen in an MS white paper on CARP that there is a
routing option called distributed routing: all requests go to one proxy
that uses the other according the CARP method. The clients actually only
see the first proxy.
Is this option available with SQUID? What is the configuration in this case?
I have seen in the SQUID doc that CARP is allowed only for parents. What
does it mean?
In case of this being not available, what is the best option to achieve
load balancing: use DNS round robbin and ICP between the 2 proxies (using
proxy-only directive in cache_peer)

2. For the branch office proxies, what could be the "hierarchical" config?
I see no interest in using ICP or CARP between these proxies and my proxy.
Is this correct (of course, in a branch office, one could use any type of
cache hierarchy but this has no impact on my proxy config, right?).

TIA

Siegfried
Received on Fri Aug 04 2000 - 07:16:50 MDT