RAID 5 v 10

[RichH]

Hi

After experiancing write queing on some production sql servers using raid5, I decided to test raid10, so I setup two identical DL380's with 6i controller one with a raid10 array for the sql data and the other with raid5.

The test was to perform a simple insert 100,000 times.

This took 6mins on both servers, exactally the same time.

Shouldn't the RAID10 have written a lot faster?

Thanks
Rich

[Jaime Bolanos Rojas.]

RichH,

You might want to take a look at this pages, on Raid 10 it specifies:

"All drives must move in parallel to proper track lowering sustained performance"

http://www.acnc.com/04_01_10.html

Regards,

Jaime.

[Ivan Ferreira]

Have you tried running other kind of test to ensure that is the disk the bottleneck?

That is, run specific disk intensive writes without other application running, to discard other variables. RAID 10 should be faster.

For disk and file system performance tools see:
http://h20000.www2.hp.com/bizsupport/TechSupport/Document.jsp?objectID=lpg50460&locale=en_US

[RichH]

Hi

Thanks for the links. I will check them out.

In the mean time, I forgot to mention I was running perfmon and on both tests the Avg Disk Write Queue was upto the 100's and processor, paging etc was pratically none.

Rich

[RichH]

I guess it could be the write cache on the controller was able to cache the requests at the same rate.

Would a longer test or larger insert be a better test. i.e. fill the cache.

Then the faster RAID10 should show its advantage?

I feel I am missing something, as it is widely talked about that RAID10 should give significant performance advantahes over 5!

Cheers
Rich