Is there an substitute for spindles?

hutchingsp

Limp Gawd
Joined
Dec 24, 2006
Messages
150
If you wanted to get absolute maximum random read performance from a disk array, given a choice, would you go for a 12x3.5" drive enclosure or a 24x2.5" drive enclosure?

The vendors recommendation is "Min. RAID 5 Raid groups with 7+ spindles each" which would then be split into several volumes before being presented to the OS.

I can't help but think that the 24x2.5" drives are the way to go as there should be no substitute for raw spindle count for this type of thing?
 
I hope you are going to use raid6 or 60 for this raid 5 is plain nuts with that many drives. Unless you do not care about your data at all.

If you wanted to get absolute maximum random read performance from a disk array, given a choice, would you go for a 12x3.5" drive enclosure or a 24x2.5" drive enclosure?

24x2.5 would be much faster and you can get 15K rpm sas drives if you want the most performance.
 
Last edited:
It would be several (4 or 6) RAID5 sets so rebuild times shouldn't be a huge issue.

The drives would be 1tb 7.2k MDL SAS. 15k are too much money and wouldn't give me enough capacity.
 
Yes more spindles is key when dealing with random reads. Hell even same RPM the 2.5 inch drives will usually be better at random disk I/O than their 3.5 inch counterparts (probably because the head doesn't have to move as much on the smaller platters).

Benchmarking various setups here is what I saw:

Seeks/sec:
SATA 883 Seeks/sec
2TB 891 Seeks/sec +0.91%
Dell 989 Seeks/sec +12.01%
BB 2521 Seeks/sec +185.50%

SATA=8x1TB 7200 rpm raid10
2TB= 8x2TB 7200 RPM hitachi sata raid6
Dell= 6x500GB SAS (2.5 inch) raid6
BB =24x3TB 7200 RPM ultrastars raid6

As you can see even though it had 2 less spindles the random seeks/sec on the 500GB 2.5 inch SAS disks were greater than 1 TB 3.5 inch disks that were SATA even though the SATA setup had two additional disks.
 
Back
Top