Return to BSD News archive
Path: sserve!newshost.anu.edu.au!munnari.oz.au!foxhound.dsto.gov.au!fang.dsto.gov.au!yoyo.aarnet.edu.au!news.adelaide.edu.au!news.cs.su.oz.au!harbinger.cc.monash.edu.au!yeshua.marcam.com!MathWorks.Com!noc.near.net!news.delphi.com!usenet From: John Dyson <dysonj@delphi.com> Newsgroups: comp.os.386bsd.questions Subject: Re: linux's I/O calls faster than NetBSD's ? Date: Fri, 11 Mar 94 01:56:19 -0500 Organization: Delphi (info@delphi.com email, 800-695-4005 voice) Lines: 19 Message-ID: <5w-r3h7.dysonj@delphi.com> References: <2lhv9r$pbt@homea.ensta.fr> <CMFvuI.IHu@cnsnews.Colorado.EDU> NNTP-Posting-Host: bos1b.delphi.com X-To: -=Runaway Daemon=- <frechett@benji.Colorado.EDU> -=Runaway Daemon=- <frechett@benji.Colorado.EDU> writes: >I'm not sure if I like unified caches yet as I've seen them seriously >cripple an Alpha box. Having not spent much time in Linux I can't >say if it's better than OSF/1's implementation or not. On FreeBSD, we will be implementing a unified cache (in progress now.) There are a couple of things that can really cause problems. We will not simply allow the buffer cache to overrun the rest of memory, and also we will not let lots of dirty buffers sandbag the disk drives. There will be a signficant performance improvement because of some of the leveraging of the VM system against the problem also. If anyone has comments on the problems or difficulties with unified caches, let me know. I do have a variable size buffer cache version of vfs__bio that I wrote about a year ago, and it could be coerced into causing problems (hope to avoid in the new unified code.) John dyson@implode.root.com