DiskMark - 10K and 15K Drives

[CEW]

I have a Dell PowerEdge 2900 with a Perc 5/i disk controller. I have two RAIDS configured:

RAID 1 = 2 x 15K SAS Drives (36GB/Each)
RAID 5 = 6 x 10K SAS Drives (146GB/Each)

I ran the DiskMark and the performance of the RAID5 seems reasonable, but the performance of the RAID1 seems rather low. Does anyone agree that the performance of the RAID1 should be better?

RAID1 DiskMark = 626
RAID5 DiskMark = 2754

I know that mirroring had extra overhead and probably won't be as fast as a RAID5 with 6 disks, but these are 15K drives so I would expect better than 626. I ran numerous tests and this is the average, though each test was within a few points of each other.

[David (PassMark)]

What were the individual numbers for each of the 3 tests?

[CEW]

Thanks for the follow-up. Here are the individual scores:

RAID1
Sequential Read = 83.4
Sequential Write = 77.5
Random Seek + RW = 12.1
DataMark = 625

RAID5
Sequential Read = 417
Sequential Write = 225
Random Seek + RW = 110
DataMark = 2718

[David (PassMark)]

I suspect the RAID5 is using a lot more disk caching. The disk test in PerformanceTest tries to minimise the impact of caching on the results. But we don't have any control over what happens in RAID controllers.

I think it might be interesting to use the advanced disk test in PerformanceTest to run a few tests with files large enough that they can't be cached (50GB+). And then see what the results are.

[CEW]

Thank you for your prompt feedback. I will take a look at the cache settings on the controller and run further tests. I will have to wait for a good time when I can reboot the server to get the cache settings. I'll post back with my findings.