HPE MSA Storage Configuration and Best Practices for VMware vSphere

[GMZ]

Hi

I've been reading the above Best Practice guide, but it's dated August 2017 Rev 2.

Is there anything newer than this relating to VSphere?

Im just a little unsure about page 25 where it says:

"VMware vSphere® VMFS datastores on multiple Virtual Volumes on the same Virtual Disk Group can have an adverse effect on overall system performance of the vSphere cluster. Using the above recommended configuration of a single Virtual Volume for each Storage Pool with a Virtual Disk Group aligned with the Power of 2 Model will maximize performance and capacity."

So is this basically meaning  1 VMFS datastore : 1 Virtual Volume : 1 Virtual Disk Group ?  If so, then it seems that Tiering is not recommended.  Or is the emphasis on creating a single Virtual Volume for each pool.

I am not actually sure how much of a benefit an Archive tier would be with VSphere.  An actual example given in the guide was of a single RAID6 Tier plus a single SSD for read cache.

Thanks,

Gordon

[SUBHAJIT KHANBARMAN_1]

As per my understanding yes this the last best practice guide release specifically based on VMWare,

https://h20195.www2.hpe.com/v2/getpdf.aspx/4aa4-7060enw.pdf

Coming to the portion that you have highlighted, this is just an ideal example but that doesn't mean everyone wil follow this.

Obviously if anyone having single pool, single Virtual Disk Group and multiple volumes created. Those volume presented as datastore in VMFS. On those datastore each having multiple VMs running then performance impact will be there compared to single pool, single Virtual Disk Group and single volume where few VMs running.

Ofcourse tiering is important. It depends on what type of drives that you are using. In ideal system you should have Performance, Standard and Archive type of drives so that respective type of VDG can be created under same Pool.

You can also check volume affinity feature and that can also help for specific purpose.

Hope this helps!
Regards
Subhajit

I am an HPE employee

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

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

[SUBHAJIT KHANBARMAN_1]

Do you have any update or outstanding query on this topic?

Request you to please update on the forum so that everyone can get the update who are all following.

If you have no further query and you are satisfied with answers then kindly mark the forum as resolved.

Hope this helps!
Regards
Subhajit

I am an HPE employee

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

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

[GMZ]

The document you referred to https://h20195.www2.hpe.com/v2/getpdf.aspx/4aa4-7060enw.pdf on page 24 (and the "Best Practice" on pg 25), HPE recommends to use a RAID 6 VDG for pools A and B, and use a single SSD drive for read cache for each of those.  So there is no tiering recommended for vSphere virtualizaton environments!

I envisage that most buyers of the MSA 2052 would be using it for virtual environments, so would these buyers be better off just buying a 2050 along with it's read cache functionality.

I'd love to hear of some real world users of the 2052 with virtual workloads and what their disk setup is like.

You mentioned "In ideal system you should have Performance, Standard and Archive type of drives", but you must be referring to non-virtual workloads right?

[SUBHAJIT KHANBARMAN_1]

MSA2050 or MSA2052 there is no concept of linear anymore. All Virtual only.

If anyone go for MSA2050 with reach cache option then that person will achive performance in terms of read operation only when using SSD drives

If anyone go for MSA2050 or MSA2052 with tier option then that person will achie performance in terms of read and write operation both when using SSD drives

Hope this helps!
Regards
Subhajit

I am an HPE employee

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

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

[GMZ]

Subhajit,

Maybe I wasn't clear...My question wasn't about the difference between read cache and tiering.

The document 4AA4-7060ENW, from August 2017, on page 24 does not recommend tiering, but instead recommends RAID 6 and a read cache (not tiering).

My question is: Today, nearly 3 years after that document was writen, does HPE recommend tiering to be used with a MSA 2052 if the workload is 100% virtual machines?

I'm in the planning stages and want to follow current best practise to avoid reconfiguring the storage.

Thanks

[SUBHAJIT KHANBARMAN_1]

As per my understanding any product user guide or Best practice guide is just a generic guidance to learn the product only but in practical no document can provide exact design which can match your business requirement.

You need to choose according to your business requirement what is best fit or it's always better to get a Solution consultant or Architect to do the design based on your business requirement.

As per the best practice whitepaper page no 24, this is just an example where it shows if someone having only SAS drives and two SSD drive then how to configure the setup which can provide good performance but if someone having capacity to buy multiple SSD drives and their business requirement is write intensive applications then obviously they will go for Performance tier instead of read cahe and limited to SAS drives only.

Hope this helps!
Regards
Subhajit

I am an HPE employee

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

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

[Shawn_K]

Hello GMZ,

Without knowing exactly how you are planning to deploy your VMs it is hard to give you specific answers to configuration. Please bear in mind, my answers will be aimed at a high level and not intended to fit every user using an MSA and VMware.

Tiering is useful for keeping certain data in areas of the pool for faster access. For example, you may be creating VMs for virtual desktops for your work force. If these are being created new each day, you will end up having a high read workload as those VMs are loaded. In this instance, the READ cache can help boost your performance. However, you may also want to create VMs and then every day take a snapshot of the VM. That snapshot would not be used often so would tend to end up as a "cold" page. Meaning the level of access is low or cold and so would be moved down to an Archive tier. This would free up your higher tier to handle requests for your VMs.

What you need to consider when looking at whether tiering is going to give you any benefits is what your environment will be used for. If you are just creating VMs with datastores that will be frequently accessed then the Read Cache will help your performance but tiering may not offer that much of a benefit. You will get a benefit from tiering if some of your storage is going to be used for data that is infrquently accessed but not on a backup level.

Hopefully this make sense and answers your query. 

If you wish to check your firmware on your system you can now do so easily by checking this link: www.hpe.com/storage/msafirmware

Also, to check the overall health of your system and any available firmware versions you can use our new MSA Health Check Tool at the following link: www.hpe.com/storage/MSAHealthCheck

Download your MSA Log File from your MSA array
Upload the MSA Log File into the MSA Health Check website
Review the results by clicking through the tabs and saving the PDF report
Links to array, enclosure, and drive firmware are provided for esay access

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.

[GMZ]

Thanks Shawn_K

Things are a little clearer, but how about this ...

Referring to that Vmware Vest Practise document yet again...

On the top of p25 it recommends to use a "single Virtual Volume for each Storage Pool".  This being the case, then tier affinity is not possible.

I was thinking of hedging my bets, by having some certainty about performance for certain VMs, by creating 2 virtual volumes (VV),  setting tier affinity for the important VV to the performance tier.  Is this a reasonable strategy? 

How critical for overall performance would having a "single Virtual Volume for each Storage Pool" be, versus having 2-3 virtual volumes per pool to allow for affinity needs?