Diskgroups with different RAID levels in same Pool on MSA 2050

[pchpe]

Hello,

On a MSA 2050, 2 diskgroups are created on Pool-A: Diskgroup-1 with RAID10 & Diskgroup-2 with RAID5.

If a volume is created on Pool-A, its data will get striped across both diskgroups, is the understanding correct? If yes, then will some data go to RAID10 and some to RAID5? If yes, thats very unusual, isnt it?

I am not able to understand it. Why does the storage even allow creating different diskgroups on a Pool?

Pls help understand. Thanks.

[SUBHAJIT KHANBARMAN_1]

@pchpe 

Yes your understanding is correct. Data or specifically 4MB pages getting evenly distributed accross all VDG in the same Pool because of Wide striping feature.

Now coming to your query why Storage allow different RAID level VDG in same pool. One use case can be lets say you have some VDG present and you are running out of space. In emergency you have only 2 drives with you. Now in Virtual array you can't add drives to same VDG to extend them like linear array. In that case you have only option to create a new VDG with those two drives and add to same pool to increase space of that Pool so that volume space can be increased and you can avoid outage situation. Like this there can be many use cases. However it's always recommended to add same number of drives, same size and same RAID VDG like existing VDGs to the same Pool. This helps in improviding performance but not mandatory.

Hope this helps!
Regards
Subhajit

I am an HPE employee

If you feel this was helpful please click the KUDOS! thumb below!

**********************************************************************

[ArunKKR]

Hi,

As Subhajit mentioned, using identical configuration (RAID type) disk groups in a tier is a best practice recommendation to obtain optimal performance. Due to the immediate pool expansion requirements when the Pool becomes full or due to the number of free slots available it might not be possible to follow best practice recommendation always while creating a disk group.

[pchpe]

Thanks Subhajit.

It would be better if HP can clearly document both of the below:

- this use case that you have mentioned

- more importantly, that the best pest practice is to use the same RAID level for all VDGs in a pool

It is not very clear and because people were used to linear storage, people continue thinking on same lines and create  VDGs in the same pool with different RAID levels thinking that volumes will reside in one of the VDGs. Ofcourse its an oversight on their part, but too difficult, infact quite confusing, for poeple to unlearn and follow this new concept of virtual storage.

Better if a warning is shown at the time of creating a new VDG in the same pool with a different RAID level. Allow but with a warning.

btw, any other use cases?

[ArunKKR]

Hi,

Creating different RAID type disk groups across different tiers should not cause any performance issues if they follow the power of 2 method.

MSA 2050 supports 3 different tiers based on the type of disks being used - Standard (SAS 10K/15Krpm drive) , Performance (SSD) and Archive (7.2K rpm)
Best practice guide recommends to create similar disk group RAID type with in a tier in a pool for optimal performance.

Please refer to pages 18 and 19 of the best practices guide.

Following power of 2 method while creating disk group is more important and we have seen significant sequential write performance improvements when this suggestion is followed.

Newer version firmware gives a  pop up message related to power of 2 method while creating a disk group.

Note: Power of 2 disk group recommendation is only applicable for RAID 5 and 6.

[pchpe]

Thanks Arun

We have only the Standard tier in use (because we have only 15k rom SAS disks). And 2 DGs have been created on Pool-A, one with RAID5 and other with RAID10.

The idea was to have Oracle data go on the RAID10 DG and hadoop big data go on the RAID5 DG. But because it is not linear storage, instead it is virtual storage, we have no control on which data goes where. Both volumes are on Pool-A and hence Oracle as well as Hadoop data goes to Pool-A "All DGs" which means some parts of Oracle data are going to a RAID10 DG and some other parts of Oracle data are going to RAID5.

That is too bad, isnt it? So I guess, we will need to do something like below:

- have all RAID5 DGs on Pool-A and have all Hadoop volumes on Pool-A

- have all RAID10 DGs on Pool-B and have all Oracle volumes on Pool-B

[ArunKKR]

Hi,

Having separate Pools will also provide the advantage of controller level load balancing.

[ArunKKR]

Hi,

Please let us know whether the information shared helped.