Path: utzoo!news-server.csri.toronto.edu!rutgers!att!emory!swrinde!elroy.jpl.nasa.gov!mahendo!wlbr!voder!pyramid!ctnews!unix386!mburg
From: mburg@unix386.Convergent.COM (Mike Burg)
Newsgroups: comp.unix.sysv386
Subject: Re: Esix verses the rest of 386-unix
Message-ID: <6772@unix386.Convergent.COM>
Date: 12 Mar 91 19:45:16 GMT
References: <293@rancor.UUCP> <1991Mar10.211104.9790@cimcor.mn.org> <294@rancor.UUCP>
Organization: Unisys/Convergent, San Jose, CA
Lines: 20

In article <294@rancor.UUCP>, bob@rancor.UUCP (Bob Willcox) writes:
# >From article <293@rancor.UUCP>, by bob@rancor.UUCP (Bob Willcox):
# >Up the value NBUF made a huge difference on this box with the
# >S51K systems. FFS didn't seem reliable on this RLL drive but the 
# >S51K stuff works great. 
# 
# Since I only use FFS, I set NBUF low (100, probably could have gone
# lower though) since I don't think they are being used.  (Does anyone
# know if the 1KB buffers are ever used if you don't have any S51K 
# filesystems?)

You can set NBUF to probably as low as 10 (ten). You cannot set it
to zero because there are portions of the kernel (namely device drivers)
which use getebuf() to retrieve a buffer to perform I/O. The ECHD driver
uses it to retrieve the VTOC/PDINFO blocks from the disk upon opening of
a drive.
-- 
----------------------------------
Michael Burg -  Unisys/Convergent Corp.  Unix Intel Platforms Division San Jose
Phone: (408) 456-5934 UUCP: uunet!pyramid!ctnews!unix386.Convergent.com!mburg