[squid-users] performance questions

we are working hard on some performance issues with our squid proxies.

as background information:

at the moment there are at about 20.000 users in 150 co-locations. every
co-location has it's own squid-proxy. the topology looks like:

        co-location => central proxy => firewall => DMZ-proxy =>
internet
           
the central proxy is a double-cpu server PIII 1GHz, 1 GB RAM, 18 GB HD,
running suse linux 8.0. we have configured three different proxies on the
central proxy (one intranet, extranet and internet)

at peak we have at about 100-150 Requests/s with 13-20 MB transfer volume.

time transfer volume (MB) Objects/Requests
11.20 13.8325986862183 5691
11.21 18.0507574081421 6150
11.22 14.2347249984741 5348
11.23 30.8025789260864 5866
11.24 13.9124889373779 4969
11.25 13.1664276123047 5122
11.26 13.3603639602661 5872
11.27 13.8448610305786 5930
11.28 19.539514541626 6692
11.29 15.2315979003906 6403

no my question:
squid is sometimes very slow although we have 4 MBit connection to the
internet.
is squid (with the current configuration and hardware) capable to handle
that
load? should we use a cluster for central proxy?

this is a snapshot from top's output

 10:49am up 25 days, 17:32, 1 user, load average: 4.18, 2.83, 2.76
563 processes: 557 sleeping, 6 running, 0 zombie, 0 stopped
CPU0 states: 84.1% user, 9.1% system, 0.0% nice, 6.2% idle
CPU1 states: 64.4% user, 17.3% system, 0.0% nice, 17.4% idle
Mem: 1029344K av, 1023648K used, 5696K free, 0K shrd, 77356K
buff
Swap: 2097136K av, 64864K used, 2032272K free 350472K
cached

  PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND
  222 www 14 0 99.4M 76M 1236 R 86.6 7.6 3072m squid
20722 www 15 0 153M 147M 1252 R 60.1 14.6 8426m squid
10585 root 17 0 1256 1256 732 R 10.1 0.1 0:06 top

thanxs

Markus Rietzler
* <rietzler_software/>
* RZF NRW
* Tel: 0211.4572-130
Received on Thu Jul 25 2002 - 02:55:40 MDT