Subscription Levels in Thin Pools

Question

I'd like to know what kind of subscription levels you folks feel comfortable with in your thin pools. Myself, I'd like to stay as close to 100% or lower as possible, but I've seen 120%!

Zikas · Accepted Answer

Well Mitch don't be afraid if you have FAST/VP.

The document that (actually i downloaded today) Mr. Code provide is very good and you have to read it/study it carefully.

It is very informative.

Now regarding the tiers and the capacity, it is ok.

It depends on what appications you are using.

Actually i believe that if you have OLTP appications the 2TB will be consumed immediatelly, but it is ok.

For the 40TB of FC, it depends on your needs and your applications.

Why don't you use and SATA or SAS as well.

As i told you, if you have Oracle databases, you can provide devices for datafiles and you can fully pre-allocate those devices.

So with that move you know in any time that you have devices that already consume the space and you won't fear.

I know what you are saying. To ave a Full Pool is a very bad thing, but believe me you can avoid it.

Also there is the option to epxand the Thin Pools with new physical disks and TDATs.

Don't be afraid and try to monitor everything and this will help you a very good reporting and monitoring tool such as SRM Suite.

If you don't have money to spend now for that Suite, then try through UniSphere to monitor your thin pools.

Some reasons as well to use FAST:

The primary benefits of FAST include:

· Improving application performance at the same cost, or providing the same application performance at lower cost. Cost is defined as: acquisition (both hardware and software), space/energy, and management expense.

· Elimination of manually tiering applications when workload characteristics change over time.

· Automating the process of identifying data that can benefit from Enterprise Flash Drives (EFDs) or that can be kept on higher-capacity, less-expensive SATA drives without impacting performance.

· Optimizing and prioritizing business applications, allowing customers to dynamically allocate storage resources within a single array configuration.

· Delivering greater flexibility in meeting different price/performance ratios throughout the lifecycle of the stored information.

The need for FAST

Due to advances in drive technology, and the need for storage consolidation, the number of drive types supported by Symmetrix arrays has grown significantly. These drives span a range of storage-service specializations and cost characteristics that differ greatly.

Several differences exist between the drive types supported by the 4

Fully Automated Storage Tiering

Implementing FAST VP for EMC Symmetrix VMAX Series Arrays Technical Notes

Symmetrix VMAX Family arrays. The primary differences are:

· Response time

· Cost per unit of storage capacity

· Cost per unit of storage request processing

Zikas · Answer

Hi Mitch,

are you using FAST VP?

If yes, what code do you have? 5875 or 5876, because there is big difference how those two codes calculate the free space on the tiers.

If yes, don't be afraid to use oversubscription.

Moreover this is the benefit of using Thin Provisioning with oversubscription and FAST VP.

Symmetrix Virtual Provisioning, also known in the industry as “thin provisioning”, is the ability to present a host and therefore an application, with more storage capacity than is physically allocated to it in the storage array. The physical storage is then allocated to the application “on-demand” as it is needed from a shared pool of storage.

For example:

The basic building blocks of Virtual Provisioning are a cache only device called a Thin Device and a Data Device that is part of a shared pool of storage called the thin pool.

Host A has been given access to a 180 Gb logical volume that was created on a thin device. The host treats the device as any other device. Except for a small amount of storage space that is allocated to a thin device when it is bound to a pool, an application on host A consumes no physical storage until it starts writing to and using storage space on the thin device. If an application on the host writes 2 Mb of data to the volume, 2 Mb of actual storage is used from the devices in the thin pool. Host B is also mapped to a thin device, which shares storage with Host A. When an application on Host B writes 60 Gb of data to the thin device 60 Gb of storage will be consumed from the thin pool.

Again, storage is only consumed as needed. All the unused storage from the pool is available to be shared.

Over-subscription is where the total capacity of all TDEVs bound to a Thin pool is greater than the aggregate capacity of all data devices in the pool. This is a normal practice but does require that a storage administrator monitor the pool to prevent a “pool full” condition.

In any case this is a long story, i am not saying that you can have as oversubscription as you wish, but if you can calculate the space that your tiers have which the pool is associated with don't be afraid.

Also if you keep in mind that another way to control the oversubscription is to use fully-preallocated device provisioned to every host.

Of course ith that way you cancel the Thin Provisioning and the potential that your VMAX have and the reason why you buy the VMAX, but it is a way to control the oversubscription.

If you provisioning space to ESX, tell to your ESX admins, that you are going to provision to them Thin devices but the admins have to use those devices via the ESX as thick in order NOT to monitor them as well the allocated and the written space on the devices, i mean to avoid extra effort by monitoring the provisioned devices.

Hope that i helped you.

Mitch_Ullman · Answer

We're on 5876 and we do have a FAST VP tier set up. How does 5875 differ in its allocation calculation? It's the code I'm used to, so I may be over-worried about things.

Mitch_Ullman · Answer

We are actually using FAST/VP. We've got it broken up between two tiers: EFD & 15kFC. The EFD tier is fairly small, only about 2TB whereas the 15k tier is roughly 40TB.

I've got the policy set at a 50/100 with default aggressiveness.

My only real fear about over provisioning is that our BU do not typically communicate that they are about to slam data into a LUN.

Shortly before I got here, about two years ago, something like that happened and a pool completely filled.

Zikas · Answer

Thank you Mr. Cody. Really i appreciate your assist. You are always eager beaver to help.

Zikas · Answer

Hi Mitch don't worry, i am not hear to scare you, but to help you.

In 5875 if you have associate the pool with two or three tiers it takes into consideration the free space of each tier separately.

In 5876, it counts the free space on all tiers that the pool is associated. It is better, it gives more air to breath.

I don't remember the best practice or recommendation if you are not using FAST VP, but i think that the 140% oversubscription it is a secure number.

But why you don't have/use FAST VP Policy.

Is there any specific reason?

codyhosterman · Answer

Allocation by FAST VP Policy was introduced in 5876 and a partial reason for this change. It can help protect you against pool exhaustion too by allocating to another pool if one in the policy full. But note this is disabled by default.

https://support.emc.com/docu31003_FAST-VP-for-Symmetrix-VMAX-Theory-and-Best-Practices-for-Planning-and-Performance.pdf?language=en_US

Zikas · Answer

Exactly Mitch.

You need the SATA or SAS as well.

Data that are not used quite often, will be demoted to that SATA tier.

In any case, in ECN you can find people that will assist you and will provide to you enough info and maybe solutions.

Mitch_Ullman · Answer

Unfortunately, the technical team was not consulted for the purchase and we were told after the fact that neither SAS nor SATA would grace the VMAX. /shrug Thanks everyone for the input!

VMAX

Subscription Levels in Thin Pools

Was this post helpful?