I would say that the SAN admins are nuts if they think that the DBAs aren’t going to worry about storage performance. There is no magic SAN dust which keeps everything running fast all of the time. If you put enough load on the array then performance will be impacted, no matter how many disks there are in the array. The physics say that eventually the little disks can only spin so quickly before they break apart.
In my opinion if they tell you not to worry about performance, then they aren’t worried about the performance, which means that they aren’t going their job.
Denny
One Response
I’m in 100% agreement with “mrdenny” backed by a real world example.
We recently migrated some production SQL server form EMC storage to a HDS AMS 2500. HDS folks sword up and down that DP Pools filled with a lot of spindles would out preform standard practices. We made some nice sized DP Pools consisting of Raid 5 (4+1) groups. Well it turns out a DP Pool lacks simultaneous writes. Say you have 10 Raid 5 groups consisting of (4+1) when the storage array writes to disk it writes in a (32 MB virtual page) on one of the Raid groups once it fills the 32MB virtual page it then moves on to a new 32 MB virtual page on another Raid group. So it basically round-robins raid groups in 32MB virtual page files.
I can safely say we are now in the process of moving our systems off the DP Pools and back to a new EMC array with standard practices of separate spindles for drives.
Nick