Way more support/reports for auto-tiering (EMC)

[#1]

From what I hear, automatic tiering is the direction a lot of people are going/have gone. I recently applied the hotfix and got the "storage tiers" page working and populated. However, the information is just a duplicate of what you have setup in the array. (# of disks, # of luns per tier) and not something I would ever need to look at. In fact, I can't see this being useful to anyone. I know not all auto-tiering works the same for all vendors, but EMC moves it around in 1gb chunks, so the # of luns per disk type is really useles because a lun can sit on ALL tiers. I was hoping to have some sort of reporting on what data was moved where. (10gb moved up from Tier 2 to Tier 1 over the last day) Or iops per tier... There are dozens of other reports that would be helpful. Or a monthly historical report on how much data was moved each night. I'm sure commenters will have more ideas to add. Anything more than what is there would be extremely helpful.

[dsbalcau]

Cross-posting this FR which covers the same Idea: http://thwack.solarwinds.com/ideas/2156

[dsbalcau]

Definitely would be interested in some more feedback about what you would find useful from this feature. Feel free to post here or email me directly. Very happy to jump on a call as well to discuss this.

[crossfiter14]

Yes, I would love to see auto-tiering reports! 

I would love to be able to run a report that gives me the source LUN of the 1gb chunck that was moved to faster/slower disk for an entire EMV VNX auto-tiering storage pool.  I would also like to see a report of all of the chunks of 1gb data that was moved to faster/slower disk in an auto-tiering pool.  In conjunction on both of these reports, I would love to see IOPS trending to give me an idea on what caused them to be moved in the first place. 

In the same vein, I would love to see better IOPS reporting in general but that is a different post for a different thread.

Looking forward to seeing this functionality being built into Storage Manager!

[dsbalcau]

Hi crossfitter14,

Thanks for your feedback. I have a couple of points regarding your comments:

1) If you had this data, what would it allow you to do? How would it make you more effective or efficient at managing your storage environment?

2) Data at this granularity simply is not available via SMI-S. You can only get sub-LUN information of this granularity with specialized EMC trace tools, AFAIK, and the load that this would cause on your array would be extremely counterproductive. (Imagine turning on debug traces for an entire operating system.)

I state these points not to be argumentative, but to get a basic understanding of what you are trying to accomplish with this data and also to get to a mutually beneficial implementation given what data are available. From EMC's perspective, this technology should just be "handling" all of these choices for you. You should not need to actively "manage" the process other than defining your Service Levels for the LUN's you have provisioned. It then immediately begged the question of why you need to know the internals of the algorithm or exactly how it's managing that data?

[crossfiter14]

Hi Balki, I completely understand why you would ask all this. 

Let me answer the questions in your first point -

To give you some background I am running this VNX in a multi-tenant environment.  The implication here is that one or more tenants can reside on a LUN depending up on how the LUN is used.  If I know details for the specific LUN and what data is moved to faster or slower disk.  For instance, this would provide me with insight on which LUNs data are continually getting moved to faster disk and I would know which tenants applications are utilizing more IOPS. 

Viewing this from an auto-tiering perspective would let me know the "heavy hitters" from a LUN perspective that need to be moved on a continual basis.  That means that I know which tenants may need a different pricing module around their storage utilization and performance.  This also means that I would know what type of data is continually being moved to provide me further insight on how to make the best decision for the tenant.  If that was possible I could create reports on a per-tenant basis to provide them a view of what data they had that was moved to slower/faster disk.

As far as your second point -

This makes complete sense as EMC's tools put a very heavy load on the array and I can only run them during pre-defined, "low use" time periods.  I thought it may have been possible to do this through SMI-S and put less of a load on the array.

All that being said, I simply am just looking for as much information about all data that is moved in the auto-tiering pool because at the end of the day I'm responsible for how the LUNs are handled even though the auto-tiering technology is handling them.  Therefore, I need to be able to provide as much data as possible. 

Does all that make sense?

Thanks for your quick reply!