Re: How much memory is enough?

On Wed, 22 Jan 1997, Serge Krashakov wrote:

> Antony Tovar <atovar@unive.it> wrote:
>
> > I am currently running Squid 1.1.0 on Linux 2.0.0, as a departmental Proxy
> > Server (60 users), on a 486 with 16MB of RAM. I need to recommend a system
> > to act as the parent cache for the entire organization. I would like
> > to use an inexpensive (and easily repaired) Intel clone, such as a
> > Pentium 133, with a 2GB SCSI hard drive, and Linux.
> >
> > My question is: how much memory is enough? The standard for most of
> > our new desktops is already 32MB. By varying the "cache_mem" setting in my
> > department's SQUID.CONF, I seem to be able to limit the memory usage to the
> > available RAM, thus avoiding the disk memory-swap-file. But I have seen
> > some messages referring to Squid's "continual increase" in memory usage,
> > over time. Does this mean Squid needs to be re-started on some sort of
> > regular basis? Is the amount of RAM used a function of proxy disk cache
> > size, number of simultaneous users, or both? Can we install this
> > simplified system (P133,32MB RAM, 2GB disk) and "force" Squid to run in
> > less than 32MB of memory? If we force it, and 32MB isn't "enough",
> > what will be the consequence, (only) slow performance or something more
> > drastic?
>
> I'm running regional cache server serving about 1500 requests per hour
> (1 per second in peak time) on P5/120-FreeBSD with 32MB RAM and 2GB disk
> without any problem (old version squid-1.0.18).

For what it's worth, I also run a department cache with about 2500
requests per hour. I run it with no problems on a P-133/64MB RAM/2GB SCSI
disk on FreeBSD-2.2. Great combo (FreeBSD and Squid).

-Mark

>
> Serge Krashakov
>
>

Mark Mayo
mark@crle.uoguelph.ca
CRLE
Received on Thu Jan 23 1997 - 15:25:00 MST