New DAE RAID best practice?

more spindles = more performance. Look at this paper for some general guidelines.

White Paper: EMC Unified Storage Best Practices for Performance and Availability: Common Platform and Block Storage 31.5 - Applied Best Practices

If you are at the latest VNX OE before you add the disks to the pool it will automatically rebalance

What does 'rebalance' mean?

Also, keeping in mind that I have 1 pool using RAID 5, when I expand the pool, will it just add the disks to the existing RAID 5 array, or will it create a new array?

Thanks,

Mike

Rebalance means it will redistribute (restripe) your existing LUNs so that they use all of the disks

Okay, so if I add my new shelf to my first pool and I already have a hot spare configured, how many 'spare' drives will I have? 1 hot spare + 1 RAID 5 = 2?

My current Pool 0 has 10 drives...does this mean that I have 1 hot spare + 1 (4+1 RAID) + 1 (4+1 RAID) for a total of 3 'spare' drives?

The only RAID Group I have is for the Hot Spare, all other disks are in Pool 0.

Mike

Not sure if I understood you properly, but if you create a second pool with 15 drives (considering you have enough spares and you don't have to make one of the drives in the new shelf as hot spare) FLARE OE will create 3 private raid groups R5 (4+1).

So if you have multiple drive failures in the same private RG (I have never seen this happen) then you are out of luck. You can still be functional with 1 drive failure in different private RGs though.

Unless you have a specific workload, which needs dedicated storage on the backend, I would recommend putting new drives in the first pool that created. If you upgrade to latest Flare OE before doing this, data will be rebalanced along all the drives on array, which will give you better performance and more flexibility.

When you say array do you mean Raid Group?

I think I'm going to create a new pool using the whole shelf. Thinking about multiple drives failing at the same time - having a small SAN I can't really tolerate multiple drive failures very well...so I'll build in a little more redundancy with another pool and RAID 5.

Mike

You have 1 hot spare (which is global in nature, meaning it will replace any failed drive throughout the array as long as its the same type and size). In a R5 (4+1) config the 1 drive represents parity (parity is distributed ofc) drive. EMC recommends 1 spare drive for every 30 drives, so you are good on spare drives.

Your current pool has 2 private raid groups and when you add the new shelf drives to Pool 0 you will have 5 total.

Again, all of the above is assuming that all these drives are of the same type.

https://community.emc.com/thread/137020

That's the part I wasn't clear on - when the Pool allocates drives it automatically splits them into 5 drive groups of 4+1...correct? That's good information since I was thinking of standard RAID when an array is created there are a number of drives and then parity accounts for a total loss of 1 drive...so a 15 drive RAID 5 array would be 14+1...but this isn't the case with a VNX Pool. In a Pool it will be 15 drives = (4+1) + (4+1) +(4+1) for a total of 12+3. Is this right?

Is there a way to see the automatically created RAID Groups? They aren't listed in the RAID Groups tab and the Properties of the Pool doesn't show anything either.

Mike

That's the part I wasn't clear on - when the Pool allocates drives it automatically splits them into 5 drive groups of 4+1...correct?

With the latest flare code, VNX will try to create either R5(8+1) or R5(4+1) so if you were to add 18 drives then 2 R5(8+1) private raid grps if 15 drives then 3 R5(4+1).

That's good information since I was thinking of standard RAID when an array is created there are a number of drives and then parity accounts for a total loss of 1 drive...so a 15 drive RAID 5 array would be 14+1...

Parity will help in case of data drive failure belonging to same raid group. So if you have 2 R5 (4+1) raid groups A and B and a drive on RG B fails then RG A parity will not be helpful.

In a Pool it will be 15 drives = (4+1) + (4+1) +(4+1) for a total of 12+3. Is this right?

That is true even in case of regular raid group based provisioning.

Okay, thanks for all your help. I went ahead and expanded my current Pool to use the new shelf and have already created LUNs and Datastores.

Thanks!

Mike

Good to know this Mike.

Would urge you to mark the thread as anwered so that we can see it as closed.

Thanks

One more question - since it was brought up in the thread. How do I know if the system has rebalanced?

Mike