Date: Fri, 12 Oct 2001 21:35:41 -0600 From: Scott Long <scott_long@btc.adaptec.com> To: Mike Smith <msmith@freebsd.org> Cc: David Greenman <dg@root.com>, Alexey Koptsevich <alex@astro.su.se>, scsi@freebsd.org Subject: Re: Adaptec 3210S: RAID5 performance and bus throughput Message-ID: <20011012213541.B464@hollin.btc.adaptec.com> In-Reply-To: <200110130053.f9D0rNk05668@mass.dis.org> References: <dg@root.com> <20011012165318.C18001@nexus.root.com> <200110130053.f9D0rNk05668@mass.dis.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Fri, Oct 12, 2001 at 05:53:23PM -0700, Mike Smith wrote: > > You only need to do two reads, one from the disk you are updating and one > > from the disk that has the parity. When you XOR the old data from the parity > > data, you've removed it and can XOR in the new data. Two reads, two writes. > > Blah. I knew my numbers looked wrong, but I was busy trying to make the > other point (you don't have to read the stripe) and instead made a fool > of myself. I wasn't very clear in my original explaination, but I meant what David and Justin said. Anyways, the point is that RADI5 writes are slow no matter how you look at it, requiring a minimum of 4 disk operations for every case that I can think of except one (write a full stripe to a 3 disk set only requires 3 operations) To get back to the original point, some cards are inherently faster than others. I don't know of any cards on the market right now that can saturate a 64/66 PCI bus, and most can't saturate a 64/33 bus. I'm not in the position to preach the pros and cons of a 3210S in public; you'll have to research that yourself. Scott To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-scsi" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20011012213541.B464>