THE SQL Server Blog Spot on the Web

Welcome to - The SQL Server blog spot on the web Sign in | |
in Search

Linchi Shea

Checking out SQL Server via empirical data points

Should I Use a Windows Striped Volume?

In Windows Server 2003, you can use the Disk Management console to create a striped volume over multiple dynamic disks (well, you can also create a mirrored, a RAID-5 volume, etc). If these disks (or LUNs) are presented from a SAN, most likely you can stripe across the same storage devices--making up these LUNs--inside the SAN to present a single LUN to Windows.

The question is: if you can create a striped set either inside SAN or at the Windows level, is there any performance advantage of doing it at the Windows level?

From what I have seen so far, you are probably better off using a SAN-based striped set rather than create a Windows striped volume. Although in no way I'm recommending this in all cases, I do have some empirical data to back myself up.

Recently, I had three 32GB LUNs presented to a server. I first used the Windows disk management tool to create a 96GB Windows striped volume on top of these three LUNs, and conducted a series of I/O performance tests on the Windows striped volume. Then, I returned the three LUNs to the SAN, and a 96GB LUN was created to stripe across the same devices that used to make up the three 32GB LUNs. Finally, the same I/O tests were conducted on the 96GB SAN striped LUN.

The I/O tests included 8K random reads, 8K sequential reads, 8K random writes, and 8K sequential writes. There was no difference in I/O throughput between the two configurations when it came to 8K random/sequential reads and 8K sequential writes. However, the SAN striped LUN significantly outperformed the Windows striped volume for 8K random writes, as shown in the following chart.

Now, how much of this 8K random write difference may translate into the performance difference in SQL Serer transaction throughput? Probably not as much or not a lot. But unless there are the numbers to show some performance advantage of using a Windows striped volume over a SAN striped LUN, why use it?

By the way, note that to create a Windows striped volume, you need to convert the LUNs to dynamic disks which are not supported in a Microsoft Cluster Server.


Published Monday, March 12, 2007 1:25 PM by Linchi Shea

Attachment(s): WindowsStriped.gif



Joe Clifford said:

Creating a stripe set via Windows disk management is definitely something I would avoid - use either the SAN or a local RAID controller to manage stripe sets as creating a stripe set via Windows is a definite performance killer as you've put the onus of managing the storage volume on the O/S rather than on hardware.  One thing to look at during your testing would be CPU utilization - my bet would be that the Windows striped volume causes a significant increase in CPU utilization when compared to the SAN volume.

One other concern with Windows striped/managed volumes is that the settings themselves are kept within the O/S - lose the system and you may very well be unable to access your data (unless you're booting from the SAN?)


March 18, 2007 12:27 PM

Linchi Shea said:

Actually, the CPU utilization in terms of % Processor Time, % Interrupt Time, and % Privileged Time was comparable in both cases: (1) Windows striped volume and (2) striped volume within the SAN.

But your point is well taken, and consistent with the message in this post.

March 21, 2007 12:59 PM

better solutions dude said:

Isn't it an overkill to have software RAID when there are enough good hardware around?

March 24, 2007 8:16 PM

Linchi Shea said:

Sometimes one may have to work with the LUNs one has, not the LUNs one wish to have :-) Thus, it is still a legit question to ask whether additional performance can be obtained by striping the LUNs at the OS level. I have heard this question asked far more often by folks on the non-Windows side. Some of them seem to favor striping smaller hardware LUNs with Veritas volume manager. I don't know whether there is any hard empirical evidence to backup that position.

March 25, 2007 5:21 PM

Tony Rogerson said:

Hi Linchi,

I wonder if the degradation is because the chunk of data being written to the Windows stripe is being split up at the Windows side and presented to the SAN as two smaller blocks of data - with the SAN stripe, the data is presented to the SAN as one block of data.

It would be interesting to test this with DASD on seperate controllers to effectively take any IOP's limits or small chunk write sizes out of the equation.

When using the Windows stripe - did you see a near doubling of IOPs on the SAN?


March 31, 2007 8:39 AM

Linchi Shea said:

> When using the Windows stripe - did you see a near doubling

> of IOPs on the SAN?

Unfortunately, I didn't check that.


March 31, 2007 10:14 PM

Bert said:

Another interesting test would be database striping.

present teh 3 columsn to the windwos OS adn then create a database with a file on each drive.


April 9, 2007 9:28 AM

Linchi Shea said:

Yeah, that is basically to test the SQL Server proportional fill algorithm as a striping method.


April 9, 2007 2:26 PM

Isaac Story said:

In response to something Joe Clifford said above, if the host that contains dynamic disk is lost, the dynamic disks and there configuration is not lost. There are signatures written on each physical disk regarding the configuration of the group that it belongs to and moving them to a new server simply involves re-mapping them over and importing the dynamic disk configuration.

My main question is when you do not have the option of striping on the back end, is third party disk management software like Veritas better then built in Windows striping? Microsoft has always plugged the practice of spreading across as many spindles as possible (especially with SQL), and striping seems to be the best option. But I need to know if there is third party software that perfoms better then built-in Windows striping.

June 21, 2007 1:16 PM

LJB said:

The built in Windows LDM and stiping is actually a 'lite' version of the Veritas (now Symantec) product.  In fact there was a law suite over it.  See:

But the question remains whether the full product is any better than the 'lite' version and it appears that hardware stiping is faster than software.

May 30, 2008 12:43 PM
New Comments to this post are disabled

About Linchi Shea

Checking out SQL Server via empirical data points

This Blog


Privacy Statement