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

No idea. Run very similar configurations (a few MHz less CPU power,
Linux-2.4.X) for about 700 requests/s in benchmarking, and the CPU usage
is quite linear up to the point where there is no CPU left..

Regards
Henrik

"S³awomir Chrószcz" wrote:
>
> 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 S³awomir
> 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
Received on Wed Jan 08 2003 - 03:18:36 MST