*BSD News Article 34411


Return to BSD News archive

Path: sserve!newshost.anu.edu.au!harbinger.cc.monash.edu.au!yeshua.marcam.com!MathWorks.Com!news2.near.net!news.delphi.com!usenet
From: John Dyson <dysonj@delphi.com>
Newsgroups: comp.os.386bsd.misc
Subject: Re: Whats wrong with Linux networking ???
Date: Tue, 16 Aug 94 09:56:23 -0500
Organization: Delphi (info@delphi.com email, 800-695-4005 voice)
Lines: 18
Message-ID: <RA2TdCP.dysonj@delphi.com>
References: <RSANDERS.94Aug9003813@hrothgar.mindspring.com> <1994Aug13.012953.5809@cs.brown.edu> <32jp2u$bpb@Starbase.NeoSoft.COM> <32ll52$n7d@quagga.ru.ac.za> <1994Aug15.034939.20997@cs.brown.edu> <32ol5s$ck@euterpe.owl.de>
NNTP-Posting-Host: bos1c.delphi.com
X-To: Martin Husemann <martin@euterpe.owl.de>

Martin Husemann <martin@euterpe.owl.de> writes:
 
>Linux uses more (upto nearly all available phys.) memory for disk cache buffers.
>NetBSD restricts this to a fraction of physical memory.
 
FreeBSD V2.0 will allow sharing of the process and buffer cache in a
very consistant fashion.  The methodology will be to use the VM object
cache mechanism, and normal file data will not reside in the buffer
cache, but in the VM cache.  We have worked out a fairly good scheme
that appears to mitigate the effects of hogging that occur with many
merged schemes.  FreeBSD will have a true merged cache -- I believe
that Linux has a buffer cache that is shared with process physical
memory.  The Linux scheme works, but so will the new FreeBSD scheme.  I
have prototype versions of the code running and I have seen no bad side
effects yet.  (I am sure that there will be some :-().
 
John
dyson@implode.root.com