Hi Imran,

Please refer the below mentioned document for general recommendations:

http://powerlink.emc.com/km/live1/en_US/Offering_Technical/White_Paper/h8222-vnx-virtual-provisioning-wp.pdf?

Also refer the below mentioned documents for more information:

http://powerlink.emc.com/km/live1/en_US/Offering_Technical/White_Paper/h8268_VNX_Block_best_practices.pdf?

Regards,

Suman Pinnamaneni

Hi Imran,

How much more storage do you need to add. Can you please share the capacity and performance calculations for sizing of new storage? It is strongly recommended to go through the internal architecture of Pools with regards to the internal private raid groups built within the pools. However in short, if you add a small number of drives to the existing pool that is almost full in capacity, then you would end up getting a structure with only the new drives which are smaller in number and you wouldn't get the same performance from the pool luns as you used to get earlier. This would be equivalent to adding another pool with the new drives. Would be glad to hear some feedback

HI Sushant,

Thanks for the suggestions.

I need to multiply existing pool by 2… means I need to double the capacity of existing production pool.

The main thing that I am concerned about is availability of pool in case of multi disk failure as existing pool is configured with RAID 5 (12 TB usable with 40 x 300 GB 15 K SAS, 5 x 1 TB SATA, 5 x 200 GB SSD ).

Pool availability is independent of the extra disks in the pool as long as you have enough disks configured as hot spares. This is what I think. Since the poos are internally made up of private raid groups with the RGs configured like the legacy RG architecture, the pool availability should not be an issue. We haven't had any disk failure in our two VNX arrays since they were configured 10 months ago. I would recommend reading up the white papers of VP and VNX Flare 31.5 Best Practices Guide. I will let you know if I find out something.

I would disagree, one private raid group going toast, the whole pool is garbage. Read about failure domains.

Right sir. Thats what has been the main limiting factor for pool expansion that the data is not going to be rebalanced(as of now, good to know its coming this year ) and hence the performance drop would be visible if the same number of drives is not added to the pool. Would you however shed some more light of the possibility of pool failure with double drive failure from different private raid groups, if any?

fault domains, it's in the BPG that you referenced earlier. I look at it this way, if i have a pool that has 120 drives and the pool is at 95% utilization, what is adding another 120 drives going to give me ? Nothing in terms of performance because currently you can't rebalance the pool, but i sure have increased my chances of complete meltdown if i do experience double drive failure. If you have EFD and SATA in the first pool and you are not adding the same type of drives to the second pool, you can argue that you can no longer take advantage of FAST.  Can't wait for multi-raid and rebalance support to be introduced this year

Thanks dynamox I would surely read up on that. But if a Hot Spare kicks in, wouldn't it keep the RG and hence as a consequence the whole pool available? Multiple disk failures in the same private RG of the pool would definitely make the whole pool unusable. Would you be able to give some pointers for the failure domains as far as pools are concerned?

multiple drive failures in different private RG ? As long as they are not in the same RG, should be a non-issue.

Would you however shed some more light of the possibility of pool failure with double drive failure from different private raid groups, if any?

The availability of any RAID group, public or private is the same.  Chapter 12, 'Storage Object Availability' in EMC Unified Storage System Fundamentals for Performance and Availability would be a good place to start.  The Fundamentals document is available on Powerlink.