Date: Wed, 12 Jul 2000 11:33:42 +0930 From: Greg Lehey <grog@lemis.com> To: Linh Pham <lplist@q.closedsrc.org> Cc: Rahul Siddharthan <rsidd@physics.iisc.ernet.in>, freebsd-questions@FreeBSD.ORG Subject: Re: Vinum/RAID-5 on IDE disks Message-ID: <20000712113342.F29642@wantadilla.lemis.com> In-Reply-To: <Pine.BSF.4.21.0007111217270.91175-100000@q.closedsrc.org> References: <20000712004443.A6134@physics.iisc.ernet.in> <Pine.BSF.4.21.0007111217270.91175-100000@q.closedsrc.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tuesday, 11 July 2000 at 12:26:57 -0700, Linh Pham wrote: > On Wed, 12 Jul 2000, Rahul Siddharthan wrote: > >> On a system with 2 IDE buses and 4 disks (2 per bus), is it a good >> idea to use RAID 5 / Vinum? I read of issues regarding performance >> and fault tolerance on linux raid when you put two disks on one bus, >> and it sounded like that should apply to freebsd too, but I can't find >> a specific mention in the vinum documentation. >> >> One problem mentioned there that if a disk goes down, it can take the >> bus and therefore the other disk down with it: it doesn't matter if >> the data on the other disk is safe and therefore the system can be >> brought back up without data loss. But is there a likelihood of the >> other disk being corrupted in such an event? >> >> If there are performance issues with slave disks, are they likely to >> be show-stopping? Any real-world experiences? >> >> Thanks, >> >> Rahul. >> >> ps the linux document was >> http://linuxdoc.org/HOWTO/Software-RAID-HOWTO-3.html > > It's always best to separate heavy I/O drives (mainly hard drives) > onto separate IDE channels. But, that doesn't always mean that you > get a whole lot more performance that way. See my last message. It frequently does. > Remember that a standard UltraDMA controller can handle up to 33MB/s > per channel, ATA/66 doubles that, and finally ATA/100 brings it up > to 100MB/s. The data transfer phase of a transaction takes about 400 µs. The latency takes about 8 ms, and you spend the time waiting for that. > Also remember that hard drives usually cannot soak up the entire ATA > channel, but some drives can. But also another note is that only one > drive can communicate on a specific IDE channel at any given > time. This might reduce the overall thoroughput, but again, it's not > always going to be the bottleneck. Admittedly, there are exceptions. But they're relatively seldom. Greg -- Finger grog@lemis.com for PGP public key See complete headers for address and phone numbers To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20000712113342.F29642>