I have been trawling the web for hours trying to decided what will be best for a new SQL Server that I'm setting up, but I'm struggling to find scenarios similar to mine.

Here's what I have:

  • 6x 10K SAS drives connected to a single Dell Perc 6i RAID controller.
  • 3x databases to be migrated to the server, each with just 5-10 users and all users primarily entering data. The databases (.mdf files?) are all less than 5GB each in size and do not grow much more than a few hundred MB /year.

n.b. Server 2008 R2 and SQL Server 2008 R2

I have read a lot about optimal raid configuration and data/file placement and as such, with what I have, what raid configuation would you reocmmend and where should I place the OS, Log Files (.ldf?), Binaries, TempDB, Data (.mdf?), and have I missed anything?. Furthermore, with the exception of the OS how and when do I specify where each of the aforementioned are placed? As you can see, I'm a bit clueless, and so appreciate your patience and understanding for a one-man-band trying his best to drag our kit into the C21.


My Current Thoughts:

RAID-1 consisting of:

  • Logical Drive C: OS, Binaries
  • Logical Drive D: TempDB, Log Files (ldf files), Windows Swap File

RAID-10 consisting of:

  • Logical Drive E: Date (mdf files)
  • 1
    I think your proposed plan seems decent. Go for it! – EEAA Jul 13 '11 at 21:15

Well, since it sounds like you are running a very low-load application, it really doesn't matter much. That said, since you have six drives and three main "classes" of data you want to store (OS, Database, Logs), to me it would make sense to create three RAID1 sets, each with two of the SAS drives. That would allow you to keep everything on different spindles so they're not competing for IO.

  • Thank you for the encouragement ErikA. Does anyone have any answers on the following part of my original question:- "how and when do I specify where each of the aforementioned [Log Files (.ldf?), Binaries, TempDB, Data (.mdf?)] are placed?" – user87581 Jul 13 '11 at 21:36

Since it is low load you could do either 2 disks RAID1 (OS) + 4 disks RAID5 to get the most space or RAID10 for better performance.

There is not really any point in seperating it into sets (except maybe for seperate OS) because you don't know that the load is going to be balanced across the 3 databases. This might leave you with several disks doing nothing while one pair is overworked. Plus the striping will already spread the IO across all the drives.

To get ultimate performance there is an argument that you should put the ldfs on a seperate set because any write will go to both the mdf and the ldf however it doesn't sound like you are pushing it to a level where you need to worry about that. Of course this has a trade off on reads, other data writes and maintainability.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.