Is it advisable to use MS defrag.exe on a RAID 5?

[Ed Zerylnick]

Does one need to use a third party product like Diskeeper to safely and effectively defrag a RAID 5? Or can you use the ordinary Microsoft defragmenter utility that comes with Windows? My server is a ProLiant and the RAID is on a P400 Smart Array controller. I am running SQL Server 2005 on the server. Budget dollars are an issue, so if MS defrag is safe and effective, we would use it, rather than paying for a higher end product that does more.

[HTH_1]

We use both Diskepper and ntdefrag on different servers here; RAID 1 and RAID 5. There is no problem at all.

[Ed Meadows · Accepted solution]

I've used Microsoft's defrag on RAID 5 for as long as I can remember and it has never caused a problem.

There are some that say that due the very nature of RAID 5, defragging doesn't help much with performance, but I don't see how it can hurt it any.

[u84823]

Since the data is physically distributed by RAID-5 itslef there is essentially little value in drfragging.

If you have the battery backup on the P400 it will be reordering operations to minimise head movement, and being SQL there will be little sequential IO anyway.

If you don't have the battery, spend the money on that, which will enable write-back caching.

[Ed Zerylnick]

A fellow named David Chernicoff published a study he did using a machine configured very similarly to mine that showed a 10 to 30% performance improvement on database queries after defrag. See http://downloads.diskeeper.com/pdf/The-Impact-Of-Disk-Fragmentation-On-Servers.pdf

[u84823]

lol report paid for by diskeeper!!!

[TTr]

The defrag tools mentioned here are filesystem based (note: file fragmentation)and they do improve filesystem performance. Note that we are talking about filesystems here and not disks. There is certain overhead in the filesystem if a file is read and put together from a few large chunks rather than from many (hundreds or thousands) small chucnks.
Now how much improvement will you realize, it is a different story. It can be significant specially if your filesystems are very active in writing (such as deleting files and creating new files all the time)You can draw your own conclusions, install a *filesystem" based io benchmark such as iozone and run a test before and after defragmenting.

[TTr]

Actually iozone is not the right tool to use because it creates its own files for testing.

[David Claypool]

Defrag is NOT contraindicated with RAID 5. While RAID 5 distributes the data across spindles, it is without any knowledge of the filesystem. Defrag is to make files within a filesystem more contiguous, so that when reading or writing there is the benefit of SEQUENTIAL disk access which typically has better throughput than RANDOM disk access.

Defrag is a good thing, regardless of RAID or JBOD, and has a side benefit of using the filesystem more efficiently (less wasted space).

[Ed Zerylnick]

Okay, so I tried MS defrag.exe on my RAID 5 based on everyone's feedback. It took only about 10 minutes to defrag the backup drive but it did not defrag the two largest database backup files that account for the majority of the space on the drive - probably because there was not sufficient free space on the drive to defrag them.
The backup of the drive with the active database data on it took much longer (I didn't stay around until it was done, so do not know how long it took). It defragged all but one data file.
System performance so far has been good.
Success!