MSA 2050

[km1986]

Hi Experts,

Kindly help me with following queries w.r.t MSA 2050.

Is it possible to create more than two pools?

What is the use of disk group? when it does have role in volume creation?

For volumes creation there is no option of selecting disk groups why?

Can I have different raid combinations in a Pool? if yes, how data distribution happens ( i use Oracle database and msa 2050 storage how do i seggerate  data if i have multiple  raids in a pool? )?

How do i check write and read memory cache?

[km1986]

small correction

What is the use of disk group when it does not have role in volume creation?

[SahanaR]

@km1986 

Hi,

>The MSA array ships with and supports both a minimum and a maximum of two controllers, each supporting one pool. A pool becomes available when a disk group is assigned to it.So you cannot create more than 2 pools

>A disk group is an aggregation of disks of the same type, using a specific RAID level that is incorporated as a component
of a pool, for the purpose of storing volume data. It hekps is creating the pool and adding you drives with raid to the pool.Volumes are made up of pool space and its data/pages is distributed across based on the affinity you select.Thats how  virtualization works.

>yes you can have different combinations and that it achieved when you create disk group in a specific pool,you get an option to select the raid type.(RAID 1,5,6,10)

To understand all the virtualization concepts and data distribution please read this virtualization guide,

https://h20195.www2.hpe.com/v2/getdocument.aspx?docname=a00087404enw

>8 GB cache per controller (4 GB read/write plus 4 GB system memory),i am not sure if there is a way customers can check.I can get that info for you.

AS of read the above link and also best practice guide https://h20195.www2.hpe.com/v2/getpdf.aspx/A00015961ENW.pdf?

Hope this helps!

Thanks

[SahanaR]

@km1986 

Hi,

I did check on your query for Cache.

This is isnt configurable and its automatically adjusted by controllers as needed.

Although you can install a read cache with SSD drives just for read cache.

Thanks

[pchpe]

Thanks.

But then why create multiple disk groups at all if writes to a volume will anyways span across all disk groups in a pool? Instead one should create just 2 disk groups - one for each storage pool, isnt it?

Also, it seems to allow multiple disk groups with different RAID levels within the same storage pool.  Am wondering, what if we create 2 diskgroups - one with RAID10 and other with RAID5 and make those both a part of same storage pool?

[SahanaR]

Hi,

CReating single disk group is not going to help you use the virtual feature of the MSA which is automated Tiering.

Automated tiering
A standout capability of virtual storage is automated sub-LUN tiering, which can increase overall system performance at a lower cost than
possible with only one class of disk drive. Automated tiering distributes the contents of a volume over multiple virtual disk groups within a
tier as well as over multiple drive classes that are grouped to form a pool. This process exploits the performance benefits of a particular drive
type, but only for the part of a volume that requires

For more info please read page 11 https://psnow.ext.hpe.com/doc/a00087404enw  

Yes you can have RAID 10 and RAID 5 in a same pool.'If you want to enchance  performance you can add SSD drives to use as Read cache/Performance tiering.

It completely depends on you of how many disk groups/tye of drives you need to create based on what is your requirement.

Please also read best practice guide I shared that might help.

Thanks

[Shawn_K]

Hello,

Think of the disk groups as the building blocks of a pool. The disk group is defined as a group of individual disks to provide a logical unit of capacity and performance which depends upon the raid level of the disk group. The MSA2050 uses a tiering algorithm, to place the hottest pages on to the fastest disk-group and move the cooler pages down to the slowest disk group within the pool as the workloads changes. This algorithm is implemented without user intervention.

When you create the disk groups try to design the Pool so the Archive disk groups will use the slower drives and then use the faster spinning disks or SSDs for performance disk groups. As you design your system consider the types of drives you have and assign a disk group and RAID level that will give you either the best performance or redundancy needed for your application and usage.

As mentioned previously, refer to the documentation to gain a clear understanding of virtual pools and disk groups. I suggest the following guides.

MSA 2050 Best Practice Guide: https://h20195.www2.hpe.com/v2/getpdf.aspx/A00015961ENW.pdf?

MSA Gen5 Virtual Storage: https://h20195.www2.hpe.com/v2/getdocument.aspx?docname=a00087404enw

Also the Stroage Information Library has the User and SMU Guides: https://techlibrary.hpe.com/us/en/storage/info-library/index.aspx?cat=msa&lang=1#.XbHFdOhKgwE

Cheers,
Shawn

I work for Hewlett Packard Enterprise. The comments in this post are my own and do not represent an official reply from HPE. No warranty or guarantees of any kind are expressed in my reply.

[km1986]

Thank you for the explanation Sahara, I will go through the provided link

[pchpe]

Thanks Shawn for your explanation on the tiering feature. How does the storage software identify which vdisk is faster and which is slower? for e.g. vdisk with 6 SAS disks in RAID10 is faster than 6 disks in RAID5 but may be slower than 10 disks in RAID5.

Is there anyway we can have control over what data is placed (written) where? for e.g. we use the storage for Oracle database. and we procure only SAS disks. So far (in P2000 & 2040) we configure it as follows:

2 disk RAID1 for REDO (odd groups)

2 disk RAID1 for REDO (even groups)

4 disk RAID10 for ARCHIVELOGS

10 disk RAID10 for DATA

We want REDO, DATA & ARCHIVE to be written to separate (physical) disks and that is because we want very low latency. If the volume that stores DATA & REDO goes to the same physical disk, then latency will get affected. Since 2050 volumes are striped across all vdisks of a storage pool, we are not able to achieve what we used to do in P2000 & 2040. Is there anyway to achieve this in 2050? OR, if you agree that our requirement is valid, then should we look at some other storage? Would appreciate your views/comments on this.

Thanks