On Wed, 16 Feb 2000, Eric J Bennett wrote:
> I'm about to configure a dedicated squid proxy server for a high bandwidth
> (155mb lan with 10mbps link to the net) high use (3000 students) network,
> what should be the primary considerations here, at the moment I'm looking
> at a PIII - 500mhz with a SCSI Raid array and 256mb ram running linux.
As mentioned earlier, since this is intended to just be a dedicated proxy I'd
lose the RAID and go with separate cache disks; in our case, I/O speed is
more important than recoverability of the cache data so high-end RAID with
redundancy isn't necessary.
Our proxy servers have dedicated 100Mbps full duplex backbone LAN switch links
with an eventual 34Mbps external link to the Internet. The proxy whose stats
I'm currently reading has registered over 3,400 unique "clients" (IP
addresses, some of which are re-used as in dial-up pools). Our proxies have a
separate Fast/Wide SCSI controller dedicated to the cache disks - I have
machines with both three (3) and six (6) spindles (7,200rpm) running fine,
though with three spindles the file system had to be tuned for space rather
speed/time in order to keep fragmentation to a manageable level (with more
spindles the operating system - Solaris/SPARC - has been able to spread things
out somewhat better before fragmentation gets out of hand).
Whether or not you have enough RAM depends on the intended cache size and
eventual average object size (a function of usage and refresh rules, not
something you can pinpoint prior to going live). See the mailing list
archives for previous discussions of this. Your CPU is most likely fine.
Cheers..
dave
Received on Thu Feb 17 2000 - 02:02:32 MST
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:51:19 MST