Re: Process size..

cache_mem only determines what Squid will try to cache. Most modern
operating systems do a reasonable job of caching filesystem blocks, which
gets most of the advantage of cache_mem (except that there's copying
overhead, because Squid can't feed it directly from its memory). I would
hesitate to go much over 64M, myself, regardless of how much memory is put
in the box, because most of the advantage of in-memory caching is going to
come with the first couple meg. Every additional meg of cache_mem brings
less advantage, and more risk of inadvertant swapping.

Later,
scott

----- Original Message -----
From: <patrickg@isyndicate.com>
To: <squid-users@ircache.net>
Sent: Wednesday, November 10, 1999 12:58 PM
Subject: RE: Process size..

> On Wed, 10 Nov 1999, Dave J Woolley wrote:
>
> > > shouldn't the cache_mem be set to something like 300Mb?
> > >
> > The FAQ, incorrectly estimates the total memory as three
> > times this figure, but postings on this list indicates that
> > the reality is that it is this this figure, plus various
> > overheads, plus a multiple of the number
> > of objects in the whole cache (including, and normally mainly,
> > the disk cache).
>
> Any guess as to what a reasonable multiplier might be then?
>
>
>
/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/
\/\
> Patrick Greenwell
> Earth is a single point of failure.
>
\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\
/\/
>
>
Received on Wed Nov 10 1999 - 15:34:22 MST