I've always been from the many spindles, small disks, JBOD, TSM s/w mirror camp. In the past, I have seen a DB performance increase, by moving the DB architecture to the above configuration.
However, like many, I have watched the threads closely and I'm seriously going to consider the RAID0 striping option with my next major implementation. Not sure which way I will be going regards the mirroring (H/W vs S/W). I've been involved in a DB corruption, caused by H/W mirroring and coupled with the fact that the previous DB backup was corrupt, it wasn't pretty. Conversely, I have been through a number of 'unscheduled' server downs, with TSM mirroring and never had a DB corruption. I have been experimenting with a Wintel server and an old Compaq FC RAID array with 12 x 18GB disks and so far the RAID0 option is coming out on top. I appreciate that it's Wintel and Compaq, but the T&R budget doesn't stretch. It's certainly a step in the right direction, a Linux TSM install & test will be my next move. Leigh -----Original Message----- From: ADSM: Dist Stor Manager [mailto:[EMAIL PROTECTED] On Behalf Of Allen S. Rout Sent: 07 February 2006 03:26 To: ADSM-L@VM.MARIST.EDU Subject: Re: [ADSM-L] Database mirroring, again >> On Mon, 6 Feb 2006 13:07:08 -0600, Roger Deschner <[EMAIL PROTECTED]> said: > Hardware 2x2 SSA RAID10 (2-way striped mirrored pairs), raw volumes, 4 > dbvols per virtual RAID volume (hdisk), which is still 2 dbvols per real > disk. I think the number volumes/spindle is increasingly irrelevant in the days of multi-GB caches in storage subsystems. My SSA is one thing, a sharkesque gizmo is quite another. > There has been some discussion of relative database corruption risk > with TSM mirroring versus hardware or OS mirroring. [...] I've been feeling myself change opinion on this, recently. The reliable-storage level here is between the log and the db, no? I think I'll be experimenting with OS- or hardware-level protection the next time we have a DB architecture question. Tho, with the DS6800, it may be a long long LONG time. :) - Allen S. Rout
