1

I am building a SharePoint 2010 farm with MS SQL 2008 R2 (STD) for the backend. The H/W is:

  • PowerEdge R710
  • 32 Gb RAM
  • 4x 146 GB Hard Drives
  • 1 additional 146 GB Hard Drive (hot Swap)

My plan is to use a second identical server (secondary) to mirror the first (primary) server.

So how best to do the array with this system (O/S, TEMPDB, System DB’s, MDF, LDF and Page File?

Any insights on this will be much appreciated.

Improve this question

4 Answers 4

Reset to default
4

Given that you only have 4 hard drives to use you've pretty much limited yourself. Setup two RAID 1 arrays, two disks each. On the first put the OS and tempdb. On the second array put the user databases. As the sharepoint server grows you'll need to add more disks, moving the logs and databases to seperate RAID arrays, and moving the tempdb database to its own RAID array.

This all depends on the number of users that will be using the system. If you only have 10-30 users that aren't using the sharepoint databases all that much it doesn't really matter how you lay things out.

Improve this answer
3
  • +100 Get more HDs. You really need more spindles. Also, MrDenny has it right with breaking out the various pieces (OS, DB, Logs, TempDB). SP has a ton of DBs, so having a separate array for each is highly impractical.
    – Chris S
    Oct 3, 2010 at 18:19
  • We have roughly 800 SP users. We four RAID volumes one for the OS and SQL executables, one for the SP MDFs, one for the SP LDFs, and one for tempdb. My preference would be to have the OS separate from SQL executables but we can't always get what we want. Response is very good. We do not use the SP workflow feature.
    – jl.
    Oct 4, 2010 at 14:15
  • It would be cheaper to get a couple of Vertex 2`s and RAID1 then. They'll mop the floor with 15K drives unless you make a big RAID10 array out of them. Sell the existing drives on eBay or something.
    – Mircea Chirea
    Oct 8, 2010 at 17:11
1

Ideally you would want to split out the individual types DBs onto their own RAID volumes (TempDB one array, LDFs a second array, MDFs a third and any high volume DBs onto their own arrays).

The problem being, if you are using DAS internally you probably will not be able to afford a RAID 10 for each one. My suggestion is that you consider the amount of volume you will be getting. SharePoint is going to have the same sort of IO patterns as you typical OLTP database.

What you really need to do is use tools like SQLIO and SQLIOSIM to stress test the IO subsystem to ensure it is performing to your needs. It is very hard for anyone to tell you what those are as you should know your user base and the project specifications. Here are a few links to get your research started. Do not cofuse SQLIO and SQLIOSIM. They do perform tests in different ways so use both.

http://technet.microsoft.com/en-us/library/cc966534.aspx

http://sqlblog.com/blogs/jonathan_kehayias/archive/2010/05/25/parsing-sqlio-output-to-excel-charts-using-regex-in-powershell.aspx

http://www.brentozar.com/archive/2006/12/dba-101-using-perfmon-for-sql-performance-tuning/

Its all about the requirements. I have seen SharePoint put on a single server with local SQL Server that had all the databases on a single volume perform just fine for SMBs with a low number of users. Test, test, test!

Improve this answer
2
  • SP has a ton of DBs (at least a dozen in a normal setup), so having a separate array for each is highly impractical.
    – Chris S
    Oct 3, 2010 at 18:20
  • I agree 100% Chris. That's not what I was suggesting, though. I was meaning more the different db types. TempDB one array, LDFs a second array, MDFs a third. I will edit my reply to be clear about that.
    – Robert Kaucher
    Oct 4, 2010 at 18:18
0

I am building a SharePoint 2010 farm with MS SQL 2008 R2 (STD) for the backend. The H/W is PowerEdge R710, 32 MB RAM, 4 X146 GB Hard Drives plus 1 additional 146 GB Hard Drive (hot Swap).

Whoever ordere that made sure you have pretty much a system totally unsuitable for a database server - because those love IO, which means hard discs.

Which is why companies such as SuperMicro sell cases that: * are 2 rack units high with 24 (!) hard disc slots * are 4 rack units high with 72 (!) hard disc slots

5 drive slots if a general purpose server. Unless you have a SAN, it is simply not suitable for higher performance db work - you can not put in enough discs for that.

So, you need an external disc cage.

Improve this answer
0

OK. Here's my revised plan.

  1. Place the Binaries, O/S. TempDB, Page files on a mirrored set (RAID 1) on the SAN.

  2. Buy an additional disk to a total of 6 disks then place the MDF and LDF on seperate arrays (RAID 5).

  3. Someone advised to buy a box with two controllers. Is this necessary?

Improve this answer

You must log in to answer this question.