VNX 5300 mix SAS drive speeds in a storage pool

So how does it know that they are different when it's in the same tier? I don't think that's accurate but someone can correct me if I'm wrong.

You should verify the amount of disks you have and compare that to the pool size. If the VNX pool states that the 2nd tier is for example 30TB, try figuring out where that number comes from. If that's because you have to add up the 15k and 10k drives you have conformation as well.

take a look at this discussion, specifically around FAST and rebalancing within the same tier

https://community.emc.com/thread/146946?start=15&tstart=0

you can see data distribution between tiers in Unisphere, don't think you can see 1g slice distribution within the same tier. I am sure support has tools to do that.

Thanks, that explains how the new drives are used fairly but is not clear on whether the IO split takes into account the different IOPS for the 10k and the 15k...

Yes it will. VNX will see the 10k and 15k drives as the same tier.

I don't recommend you place the 10k and 15k SAS drives in the same pool due to the fact that the VNX will place them in the same tier.

I wonder if the VNX sees the 15k and 10k drives as the same tier.

RRR, it sounds like that cmrayer already knows that it will be in the same tier. However, someone told him that the 10k and 15k drives are still going to be handeled differently. I've never heard that and don't see how that is possible.

In Unisphere they are in the same tier but the engineer did suggest that it knows they are different and treats them so, but I wanted a confirmation.

Ernes Taljic wrote: cmrayer, were you provided with any documentation by any chance?

Unfortunately not...

Correct, but move around the different tiers. Since the 146GB and 600GB drives will be in the same tier FAST will not be doing that. Now this might have changed in Flare 32, not sure.

cmrayer, were you provided with any documentation by any chance?

But now FAST supposedly will keep moving slices around to keep performance equal.

If you take a 15k 146GB drive and a 15k 600GB drive, the 146GB drive will outperform the 600GB drive even though both are 15k drives. I need to find documentation that backs this up; I read this somewhere but cannot find the doc. It basically talked about the arm not traveling as far as on the 146GB drive and thus the seek distance is reduced. Now how much is it really being reduced, who knows. Maybe that's why they have the capacity statement in there.

Message was edited by: Ernes Taljic. Replaced latency with seek distance

Thanks Ernes for trying to find it, I guess we'll just wait!

it makes sense that you would not want to mix drives with different RPM but different capacity ? I don't think that would be valid anymore with code 32 that can rebalance within the same tier.