Date: Fri, 28 Feb 2003 09:19:10 -0000 From: "Paul Robinson" <paul@iconoplex.co.uk> To: "David Schultz" <das@FreeBSD.ORG> Cc: <hiten@unixdaemons.com>, <hackers@FreeBSD.ORG> Subject: RE: Disk scheduling in FreeBSD Message-ID: <IPEDKJGCDFHOPEFKLIDHOEAHCDAA.paul@iconoplex.co.uk> In-Reply-To: <20030227193119.GA37891@HAL9000.homeunix.com>
next in thread | previous in thread | raw e-mail | index | archive | help
David Schultz wrote: > The original anticipatory scheduler implementation was done for > FreeBSD 4.3. See > > http://www.cs.rice.edu/~ssiyer/r/antsched/ Yeah, I managed to grab that 45 seconds after sending my original post. I've also contacted Sitaram Iyer directly to see how he feels about this heading into BSD. He seems happy with the idea, and apparently Kirk McKuisck and Poul-Hamming Kamp have mailed him about it in the past. He likes the idea of it going under BSD license, but hasn't had time/energy to get it finished off and incorporated. His precise words were "Help is greatly appreciated :-)", so help is what he'll jolly well get! :-) > I haven't looked carefully at it myself; I've only heard about it > through someone who knows Margo Seltzer. It looks very > interesting, although I worry that it would interfere with the > prefetching and sequential heuristics that the VM system already > does (which are already VERY good[1]). Maybe the authors already > thought of that issue and addressed it. So in short, I don't know > much about it, but I think it's worth looking into. Well, I'm just a hanger-on without a commit bit, so I'll work on making it production ready in the next few weeks, post up a patch and if somebody wants to commit it, great. At the moment it's all based on 4.3-RELEASE and isn't really production ready. It does look worth doing though. > [1] The fact that the improvement Linux got out of the new > approach was so much greater than the improvement FreeBSD 4.3 > got out of it is a testament to that fact, I think. The improvements he got out of FBSD are not to be sniffed at. The paper is talking about upto 60% performance increase in certain conditions. In addition, I think the algorithms used can be tweaked to take into account how FBSD works, and some of the VM stuff can be tweaked to work even better with this. Or at least, that was my intial thought last night when I was reading it on the laptop. I know FBSD already rocks on I/O generally, but I am not adverse to look at producing patches to make it better. :-) It's going to be the week after next before I start working on this proper, but I'll let -hackers know when the patches against -CURRENT are ready. -- Paul Robinson To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?IPEDKJGCDFHOPEFKLIDHOEAHCDAA.paul>