Date: Wed, 25 Oct 2006 22:12:18 +0200 From: Willem Jan Withagen <wjw@digiware.nl> To: Guido van Rooij <guido@gvr.org> Cc: freebsd-stable@freebsd.org Subject: Re: Gmirror performanc (was Re: Gmirror question) Message-ID: <453FC522.1070200@digiware.nl> In-Reply-To: <20061025132455.GA52157@gvr.gvr.org> References: <20061025100759.GA50625@gvr.gvr.org> <20061025101801.GE23885@rink.nu> <20061025103905.GB50937@gvr.gvr.org> <20061025104829.GA41873@hugo10.ka.punkt.de> <20061025132455.GA52157@gvr.gvr.org>
next in thread | previous in thread | raw e-mail | index | archive | help
Guido van Rooij wrote: > Anyway, I created a gm device and a partition. Now the read performance > is not what I'd expect. > I have the partition on two SATA devices on different controlers. > I get around 60MB/s for each disk. I can get that speed from both disks > simultaneously. > Now when I dd from the gm device, I don't get any speed higher than that. > I tried with -b split -s <various sizes>, -b round-robin, -b load. > (dd-ing as done with a bs of 1m; I see the transaction size is 128Kb, > unless the split method is used, in which case the transaction size > gies down. When round-robin is used, the transaction size is 128Kb/s, > but the number of transaction per second goes down.). > > I cannot explain why I should not get a higher read speed. Anyone? Hee Guido, I've once ran several of these types of tests for some of the disks I collected over time. Even wrote a page on that topic. More or less as a consequence of a paper you mailed me a while ago on NFS performance :) If you want: The narative on this (don't dare calling it an article.) http://www.tegenbosch28.nl/FreeBSD/Performance/Raw-disk/ And I've compared a WD800 SATA disk with a gmirror of 2 the same disks. http://www.tegenbosch28.nl/FreeBSD/Performance/Raw-disk/wd800-sata/ http://www.tegenbosch28.nl/FreeBSD/Performance/Raw-disk/wd800-sata/ The fact that these are not mentioned in the article is that I never got around into looking why the graphs look the way that they look. Especially the write ones need some serious consideration. If you want the scripts, for some DIY: Just give me a buzz... Running this for a large disk takes al long time (>1 day) I'm currently running it on a 250Gb disk. Probably there'll be flack from people telling you not to use dd for disk benchmarking. Given its simple approach it does let you understand what you are doing. --WjW
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?453FC522.1070200>