Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!sdd.hp.com!decwrl!sgi!yohn@tumult.asd.sgi.com
From: yohn@tumult.asd.sgi.com (Mike Thompson)
Newsgroups: comp.sys.sgi
Subject: Re: very fast /tmp and /usr/tmp partitions.
Message-ID: <66584@sgi.sgi.com>
Date: 10 Aug 90 02:24:31 GMT
References: <9008081842.AA24270@mcirps2.med.nyu.edu>
Sender: yohn@tumult.asd.sgi.com
Organization: Silicon Graphics, Inc., Mountain View, CA
Lines: 19

In article <9008081842.AA24270@mcirps2.med.nyu.edu>, karron@MCIRPS2.MED.NYU.EDU writes:
> How would you configure a system to use a lot (>32 mb) core as file buffer
> space ?
> 
> How can you measure the performance improvement/degradation ?

Upgrade to Release 3.0 when it becomes available.  It has a flexible
file buffer cache that uses whatever free memory is available, without
having to do any system configuration.  The cache shrinks as user
processes demand memory.

You can reconfigure a 3.2 buffer cache by upping NBUF in
/usr/sysgen/master.d/kernel and regenerating the kernel (via lboot).
It is not wise to make NBUF any bigger than 400, which will give
you an effective buffer cache of between 1.5 Mb and 6.4 Mb, depending
on your instantaneous use.  (The buffers manage varying amounts of
memory, depending on the underlying operations.)

Mike Thompson