>
> Tom Fischer <tfischer@panoramix.rain.fr> asks:
>
> > Has anyone done any qualitative testing between systems to determine
> > if there is an optimal system to be used as a squid caching proxy
> > (DEC vs IBM vs SGI vs pentium, etc.). I've seen that the US
> > (as well as RENATER here in France) are using DEC Alpha systems (does
> > anyone know why?).
>
> It depends on load of your cache server.
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Of course it is.
>
> We uses for regional cache server in Chernogolovka (Scientific Center in
> Chernogolovka campus, ~ 1000 computers, mean 1500-2000 requests per hour,
> 3000-4000 request/hour in peak time) Pentium/120 with 32MB RAM, 2GB HDD
> (we suppose to increase disk space in the nearest future) and FreeBSD.
> And thys system works fine.
Cache in KAIST, ~700 IP addresses and 2 neighbor caches,
200,000 requests/day, 3GB/day, 13000-14000 requests/hour in peak time,
average access time when hit occurs : 1 - 2 sec/requests
Pentium 166MHz with 128MB RAM, 8GB HDD, Linux 2.0.7 box.
When we tested with 64MB RAM before, average access time was
7 - 10 sec/requests. It works fine after upgrading memory to 128MB.
>
> Serge Krashakov, Administrator of Chg-FREEnet
>
>
No Swapping!!!
Hyunchul Kim
- hckim@cosmos.kaist.ac.kr
Received on Thu Oct 24 1996 - 21:44:28 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:33:21 MST