HP/Compaq RAID-ADG (Advanced Data Guard) vs RAID 0 | SQL Server Performance Forums

SQL Server Performance Forum – Threads Archive

HP/Compaq RAID-ADG (Advanced Data Guard) vs RAID 0

HP/Compaq RAID-ADG (Advanced Data Guard) vs RAID 0+1 We currently have our 75GB SQL Server database on 5-36GB drives configured with HP/Compaq RAID-ADG (Advanced Data Guard) for a total of 108GB of usable space. I know that the DB will run faster on 2-134GB drives configured in Raid 0+1, but I#%92m not sure that the fault-tolerance of having the DB spread over 2 physical drives is safer than the current 5 drives. The main advantage of the RAID-ADG seems to be that 2 drives can fail without losing data, while RAID 0+1 is guaranteed to lose the database if both drives fail. Any comments are welcome

ADG maintains 2 sets of parity data across drives, and hence is slowest of all configurations. What I’d prefer is configuring 4 drives in RAID 5 and the 5th drive as online spare. This way you are safe even if 2 drives fail at a time. Better still, configure 4 drives as RAID 1+0 and the 5th as online spare.
I know that the DB will run faster on 2-134GB drives configured in Raid 0+1.<br /><br />No it won’t because this is impossible. That configuration requires 4 drives. <img src=’/community/emoticons/emotion-1.gif’ alt=’:)‘ /><br />You probably mean RAID 1. In that scenario you can lose 1 drive. Either way, you’d be hard pressed to have 2 big drives beat 5 small ones.<br /><br />MeanOldDBA<br />[email protected]<br /><br />When life gives you a lemon, fire the DBA.
Derrick
Our sysadmin said she has 2 134GB drives configured with raid 0+1. Are you saying that she is mistaken because the minimum number of drives required for 0+1 is 4? Is 1+0 different than 0+1?
Yes. She is very mistaken. 1. There is no 0+1 2 disk configuration.
http://www.storagereview.com/guide2000/ref/hdd/perf/raid/levels/multLevel01.html 2. There is no 134GB drives on the market unless someone created one while I was sleeping. MeanOldDBA
[email protected] When life gives you a lemon, fire the DBA.
i think rfalor’s sysadmin means a 146GB (decimal) drive, which 135.97GB binary, then after the RAID controller overhead, is probably 134GB.
in any case, if only 2 disk drives, the technically correct term is RAID 1,
when more, then RAID 1+0, in theory there is a difference between 1+0 and 0+1, but almost all vendors implement the more sensible 1+0,
there is a variation on 1+0 that allows odd number of drives. still, disk performance is achieved by distributing load across a large number of spindles (disks),
I am just curious to know if the HP/Compaq RAID-ADG (Advanced Data Guard) will pick arbitary number of disks for LUN assignment? IBM does have something similiar silly algoritham. Its for a 40 disk pick 10 random? Do you have anything for 36?
picking up on the difference between raid 0+1 and raid 1+0… raid 1+0 takes pairs of mirrored disks and then combine these mirrored pairs into a raid 0 volume. This is cool because you can lose one disk out of every pair of drives. raid 0+1 takes two raid 0 volumes and mirrors them together. This is not so cool because you can only lose drives from one of the raid 0 volumes. if you lost one drive in each volume then the whole raid set is history also note that a 4 drive raid set with a hot spare does NOT really guard against dual disk failure. It just lessens the chance by rebuilding the failed disk using the spare as quickly as possible. Cheers
Twan
I tested ADG a year or so ago and would have flaky problems with disks showing up as bad (red failure light on) and then going back into non-failure. Once I deleted the array and went with one of the standards (RAID1 or 5), the problem went away. I never investigated it further because I just wanted to test ADG for the heck of it, but it did sour my taste for it… I also agree with derrickleggett, the more drives the better. One more note, if you have a Raid5 with an online spare and lose 2 drives at once, you lose the array. The online spare takes time to replace the first failed drive. If the spare hasn’t fully joined the raid set and another one fails, you lose data.
]]>