views:

42

answers:

3

Just got a win 2003 (64-bit) server box with 6 x 320 GB disks and intending to install mssql 2005 (64-bit standard). Thinking of setting up the following:

  • Disk 1 & 2 run Raid 1, split into C:\ for OS (~50GB) and D:\ for all .ldf files (~250GB)
  • Disk 3 to 6 run Raid 0, split into E:\ for .mdf files (~700GB) and F:\ for running backups etc (~200GB)

would this be optimal?

i'm wondering if tempdb should be treated like the rest of the .mdf and.ldf files or shifted out to F:\

also googled around and abit confused about filegroups, some recommended shifting out index to different filegroups and using partitioning. would this be useful since the hardware combines disk 3 to 6 into "1 physical disk" and sql will not gain from concurrent reads?

A: 

Outside of the OS raid array, the optimal hard drive setup for SQL Server is to have two separate RAID 5 (or 10 if supported) arrays. One for data, the other for transaction logs.

The rationale between the two arrays is that, ideally you don't want updates to either to be impacted because of waiting for the other.

OMG Ponies
Correct me if i'm wrong but Raid 5 isn't as good for write intensive log files as Raid 1? Also Raid 10 uses 4 disks, so i wouldn't be able to get two seperate arrays?
jj
Fast isn't much good if there's no backup, and RAID 1 is mirroring not striping.
OMG Ponies
+1  A: 

Why would you use Raid 0 for your data files? A single disk failure would bring down your array. RAID 10 is where you need to go for this: performance + fault tolerance.

phoebus
point noted, that would mean Disk 3 to 6 running Raid 10 and i'll most probably only have E:\ (~600GB) for data files only
jj
A: 
Jack Knows Jack