Overhead time per request

Hi,
I was trying to figure out the avarage time a request remains unserved
in the queue before Squid starts processing.
The log file produced by Squid is not very useful since the elapsed time in
the access log is the whole time for a request to be completed and
cannot be used to have an idea of the machine load since it also depends
on the bandwidth and on the objects size.
I'm writing a program that sends ICP queries to the cache writing down to
a log file the time for a response to come back. The icp query has been choosen
so that the proxy always returns an UDP_MISS.

Can ICP MISS processing time be used to as a proxy/cache load metric ??

How would you do to find out if that overhead time per request
(in servers with high traffic) exceed the advantages of using a proxy ?

Nico.

-- 
Nico Tranquilli
CINECA - Inter University Computing Centre
Networks and Distributed Systems
work phone: +39 (51) 6171519
email: nico@cineca.it