Re: Squid on Digital Unix 4.0B system

On Thu, 12 Feb 1998, Nick O'Brien wrote:

> DEC Alphaserver 1000 4/200
> Digital Unix 4.0B with Patch Kit (including network performance patches)
> 192Mb RAM
> 4Gb cache on a single UFS partition on its own RZ29B disk.
> 256Mb swap space
>
> This system will be dedicated to just running Squid. To give you some idea
> of the load on our current service Squid handles around 200,000 requests
> per day (according to a line count of our last few access.log files).
>
> Is this configuration suitable for our needs?

Nick,
We're running a standard (VM) version of squid on a less powerful box than
yours, and it's handling 400,000 TCP queries a day with little apparent
strain. Ours is an Alpha server 200/255 with 4Gb disk and 128Mb of RAM.
The only slight difference is that our 4GB is split between two RZ29s,
each of which is partitioned in half. I'm planning to use all four
partitions (our LRU time is down to about 24 hours, which is silly, but
the hit rate is still 50% or so) once I can persuade someone to buy me
more memory. If I can find a *really* generous benefactor, I'm hoping to
upgrade to fast-wide SCSI one day.

> We plan to use the latest stable version of Squid. Should we use Squid VM
> or NOVM? (at the moment we use NOVM on our current server)
> Should we run a DNS slave server (pointing to our site's primary) as we do
> on the current server, or use a caching DNS server?

We've had no problems with the standard out-of-the-box setup, but if
you've had good results with something different, then probably stick with
it.

Andrew

--
       +-----------------------------------------------------------------+
       | Andrew Cormack               | email: cormack@cardiff.ac.uk     |
       | Information Systems Support, | snail: 40/41 Park Place, Cardiff |
       |    postmaster and webmaster  |        CF1 3BB                   |
       | University of Wales, Cardiff |                                  |
       | Web page and PGP key: http://www.cf.ac.uk/People/cormack.html   |
       +-----------------------------------------------------------------+