MSA 2050 disk group configuration

[fvargiu]

Hello

we are configuring an MSA 2050

we want setup both controller array in this way:

1 DG with 7 disks of 600 GB 10k each

1 DG with 3 disks of 1.8 TB 10k each

the enclusure:

1 DG with 11 disks of 1.8 TB 10k each

Is it possible to configure the MSA as reported above ?

Are there any performance issue or unpredictable error ?

Are there any impact on volume side ? for example we are forced to create LUN on a specific DG or we can create LUN without care of the DG and can be considered as 1 space with all DGs.

In case of fault of 1 controller how this configuration will impact the data on the storage?

I'm new on storage so thanks in advance for your availability and answer

[JonPaul]

@fvargiu 
Here are a couple of documents which can help in getting in best practices:
https://www.hpe.com/psnow/doc/a00015961enw
https://www.hpe.com/psnow/doc/a00087404enw

Volumes(LUNs) are created from storage in a Pool
A Pool is a grouping of Disk Groups (DG)
DGs in the same Tier will be used equally. If the DGs are different in performance or capacity you could experience unpreditable performance.
With the number of drives that you have it would likely be recommended to have a single Pool but with having both 600GB and 1.8TB you should consider 2 Pools to separate those drive capacities.
DGs which do not conform to the Power Of 2 (Po2) can have degraded performance in sequential Write workloads (Best practices).

Make sure your system is up to date:  https://www.hpe.com/storage/MSAFirmware
And do a healthcheck of the system:  https://www.hpe.com/storage/MSAHealthCheck

[fvargiu]

Hi ! thanks for your reply!

it is my bad I forgot to explain that the configuration we want to do is the same on both controller ,

so the number of disk and disk group is multiplied by 2,consider also we will add spare disks and ssd for fast cache,

we want to create multiple disk group because we have some 600 GB 10 K disk to reuse from another storage MSA of the same model, 

from documentation I did not understand if we can create multiple DG of different type of disk ,

at the moment we have not  tier license  ,

from your answer I understood that without tier license we have one tier and the disk should be of the same type and capacity , Is it correct ?

thanks in advance

[JonPaul]

@fvargiu 
Without an Advanced Data Services (ADS) license you will be able to create either SSD capacity disk-groups or HDD disk-groups and READ-CACHE SSD disk-groups.  (Capacity disk-group is one which adds capacity to the Pool, a READ-CACHE does not add capacity to the pool)

You can create multiple HDD disk groups without an ADS license, up to the maximum number of disk-groups per Pool.  Those could be Standard Tier (10k and 15k) or Archive Tier (7.2k).

All data is written to storage in 4MB pages.  When there is more than 1 DG in the same Pool, same tier (i.e. 6x 600GB in RAID 6 and 4x 1.8TB in RAID 6). The system will put one page on the 600GB DG and the next page on the 1.8TB DG.  The DGs will have different performance characteristics which will lead to "lumpy" or inconsistent performance.  Once enough data is placed on the system that the 600GB DG is full then all new data will go to the 1.8TB DG.  This will lead to imbalance with more data on the 1.8TB DG and more activity on that DG and therefore likely lower performance.
It would be recommended to put all your 600GB drives into one Pool and all the 1.8TB drives into the other and to make the DGs following the Po2 configuration and the same in RAID type and number of drives.  This will lead to the most predictable performance.
These are the recommendation as stated in the Best Practices guide.  These are recommendations and not hard requirements but if you create a RAID 1 DG and a 10x drive RAID 6 DG in the same Pool/Tier your performance will be limited often by the speed of the 2 drives in the RAID 1.

[Thomas Martin]

All MSA Systems are limited to 2 cools. Each pool can contain multiple disk groups. I would put the 600G HDDs in a disk group and assign it to pool A, for example (Raid5 or Raid6 + 1 Hotspare). The 1.8TB HDDs are placed in a 2nd disk group and then in Pool B.. I would prefer RAID MSA DP+ for these drives. You can add SSDs to each pool for ReadCache. Each controller holds one Pool. Both controllers are active in this configuration and you have the best performance. If one controller fails, the pool is switched over to the second controller. The remaining controller must then process the I/Os for both pools. After the failed controller is replaced the pools are distributed over the controllers again.

[fvargiu]

Hi

thanks for your reply !

based on the informations provided we decided to create  2 Disk Group per pool:

pool A

2 Global Hot Spare

1 DG with 9  (8+1 so we should be in the Po2 best practices)  disks of 1.8 TB

1 DG with 9  (8+1 so we should be in the Po2 best practices)  disks of 1.8 TB

1 DG for fast read cache  with 1 or 2 SSD

pool B

2 Global Hot Spare

1 DG with 9  (8+1 so we should be in the Po2 best practices)  disks of 1.8 TB

1 DG with 9  (8+1 so we should be in the Po2 best practices)  disks of 1.8 TB

1 DG for fast read cache  with 1 or 2 SSD

Is this configuration good in your opinion ?

thanks  in advance

[fvargiu]

@Thomas Martin sorry I forgot to say that DG with 1,8 TB disk are in RAID 5

[fvargiu]

@JonPaul hello

Is this configuration good in terms of permance and balance ?

pool A

2 Global Hot Spare

1 DG with 9 (8+1 so we should be in the Po2 best practices) disks of 1.8 TB

1 DG with 9 (8+1 so we should be in the Po2 best practices) disks of 1.8 TB

1 DG for fast read cache with 1 or 2 SSD

pool B

2 Global Hot Spare

1 DG with 9 (8+1 so we should be in the Po2 best practices) disks of 1.8 TB

1 DG with 9 (8+1 so we should be in the Po2 best practices) disks of 1.8 TB

1 DG for fast read cache with 1 or 2 SSD

[JonPaul · Accepted solution]

@fvargiu 
Yes that configuration will be good for performance and for redundancy.
It appears that you reviewed the Best Practices about READ CACHE with a maximum of 2 drives and maximum capacity of 4TB (per pool)
HPE recommends RAID 6 for all spinning media but much of that recommendation is based on spare availability and attention to the system health.
You might want to consider RAID 6 as the drives are already available in the system and it adds time for recovery in the case of a disk fault.  There is a performance impact but it is typically minor compared to RAID 5. 

Once you have the system configured, please run a HealthCheck on the system:  https://www.hpe.com/storage/MSAHealthCheck
This will validate many best practices and firmware levels.

[fvargiu]

Hi

thanks for your help !