Maximum number of physical disks in one Thin Pool

As per page # 16 from https://support.emc.com/docu7383_Best-Practices-for-Fast,-Simple,-Capacity-Allocation-with-EMC-Symmetrix-Virtual-Provisioning.pdf?language=en_US

"“There is no limit to the number of thin devices that can be bound to a thin pool or data devices that can be added to a thin pool.

The limit to the number of thin and data devices that can be configured within a Symmetrix system is 64,000”

There is no limit for performance, but availability may become a concern with enough disks, depending on the RAID protection, the size of the drive, speed of the drive and if you have RDF protection or not.

Hi Allain,

keep in mind that the Best Practice says for one Thin Pool that the TDAT must be from the same disk technology.

Performance wised, try to create the TDAT as bigger as you can and 8 TDAT pre physical disk.

No limitation. The number of the 64000 devices is not so much possible to catch it, but if i remember well this is the limitation.

I understand the code does not limit the number of disks.

It may be "common sense" to make several pools instead of only one, very big, with all disks of the same technology in the box.

I found a old document with recommendations for 300GB-Raid5 (limit of 256 drives), and 600GB-Raid5 (limit of 64 drives). Does this sound realistic, and can it be extended to other kinds of disks and protections?

Hello Alain and Happy New Year,

well as i said and as it is said there is no number limit of disks that can be used for TDAT.

But..... always there is a BUT. But we have to keep in mind the "double fault disk" even if you have a pool of hot spares.

So for disk group of 256 disks RAID1/0 the 256 disks is certain a good amount of disks and you can increase it to 300 disks, because it is RAID1/0.

If you want to create a RAID5 disk group again the 256 disks are good amount of disks.

Performance wised don't use 600GB disk.

Zikas wrote: Performance wised don't use 600GB disk.

i would not make these statements without understanding the workload, just because i can place more data on 600G does not mean that i have doubled the number of active data.

256 FC drives with RAID5 may be too many for good availability, but fine with RDF/S.

As to the 600GB drives, they are fine given you have enough of them for the workload.  They are generally not recommended in a FAST VP environment as they don't generally have enough spindles configured to provide the performance needed in the middle tier.  You could configure more capacity and spindles, but then you start to lose the cost benefits of FAST VP.

Those OLTP databases do they have SRDF/S? Also how many IOPS do those databases produces? If they are not demanding applications and if the SRDF does not affect the performance then for sure less than 10ms response time is very good for 600GB. Also if you use SRDF/S the GSW is on? If yes, creates/adds more that 1.5ms delay to performance.

Hello Dynamox again. You are right because i wasn't so comprehensive in my statement. As Quincy mentioned below (and my mistake i didn't spell out my thoughts) it is not recommended in a FAST VP environment and Quincy spell out why. I will try to document my statements better.

The more raid groups you put in a pool, the more chances you may encounter a dual disk failure. 

For example if you have one pair of dice, you have some odds of rolling two ones.  If you roll two pairs, you have better odds of getting at least one "snake eyes".  Even more increases the odds.

Larger drives increase the rebuild time, also increasing the odds.

The 600GB drives can be fine in a FAST-VP configuration.  It depends on your workload and your IO skew and how much capacity is in your various tiers.

Hello Quincy, Happy New Year to you and to your family.  At least from my experience 256 FC disk drives RAID5 (3+1) are "marginal" for good availability. But 256 FC disk drives RAID5 (7+1) you are in big trouble

JA,

sure, i am painfully aware of the odds having experienced dual drive failure in the same raid group a few months ago (CX3). From performance perspective how do you scale your 2nd tier pool and keep it reasonably sized.

we are using 600G with FAST-VP, performance is very acceptable (less than 10ms response times for multiple OLTP databases, some VMware datastores)

why ?

Great discussion, thank you for the clarification.