|
|
|
Supreme Being
Group: Moderators
Last Login: Yesterday @ 3:37 PM
Posts: 108,
Visits: 375
|
|
Hmm, then I'd agree that 1000 random IOPS (512b, 50/50 read/write, Q32) is a bit low.
My math would say 2000 would be more like it.
I can't tell why yours are low though from the information provided. The math we use to predict this does not account for all factors and makes some assumptions.
Something that does come to mind though... Are your 16 disks all in one large RAID 5 set? So only one parity drive?
That hurts IOPS performance too. It is probably better to do 2 RAID 5 sets of 8 disks each and stripe across them. In other words do RAID50, not just 5.
This is what we do in our physical SAN platforms with 12 or more disks and it makes a significant difference in random IOPS.
Adam C
Product Manager
LeftHand Networks
|
|
|
|