[squid-users] 70% cpu load on squid without cache_dir ???

Hi!

On our gateway I setup transparent proxy. We dont want to store any
files on local HD, only redirect WWW load to another proxy.

We use: "cache_dir null /tmp" to disable storage files on HD, I add
"no_cache deny all" for sure. I want some domains to handle directly:
   always_direct allow rsi
   always_direct allow onetwp

all others we forward:
   cache_peer OUR_MASTER_PROXY parent 8080 3130

I disable all logging:
  cache_access_log /dev/null
  cache_store_log none

We use some around 10 simple access list, and delay pools.

Our proxy handle about 30-40 request per second, and linux top
shows form 40 to 70% CPU used by squid process.

I think that is very much for process which no HD operations.
All squid work is get request and decide "handle it directly
or send to parrent", no logging, no store... So why we have 70% load?

Any suggestions?

-- 
Chrószcz Sławomir
Rybnet
System:
Celeron 1200Mhz/512MB RAM/8xNIC adapters - doing routing, and NAT
for about 500 users, linux kernel 2.2.23, CPU load (without squid) near 0.01
SQUID: 2.5 stable1 + all patches
./configure \
  --prefix=/squid \
  --enable-removal-policies \
  --enable-delay-pools \
  --enable-storeio=ufs,null \
  --enable-gnuregex