I want to create 350 gb thin meta device, with the standard tdev sizes of (17, 34, 68, 134)gb..

In  my environment, if a tdev is made that should be one of these sizes,  17, 34, 68, 134gb, additionally if there need more than the 134gb we should go for metas...so u say if it is a 21-way meta or 64-way meta it doesn't matter in thin world ?

There really is little difference since you are talking about concatenated meta volumes.

Why have "standard" size TDEVs?  In a traditional thick world, there was overhead to create volumes on the drives, and considerations to be able to utilize all the capacity.  With a cache only device, there is no concern for either of these.  If you don't need a device larger than 240GB, just make the TDEV the size you need.

However, striped meta volumes can add performance to the TDEV device, and are recommended for volumes that demand high performance.

Why should the TDEVs be a certain size?  I don't understand the desire to have them created with a standard size.

64 members is a lot.  I like powers of two for member counts, but in the thin world it isn't important as it was in the thick world.

But as I said, if they are concatenated, the IO is going to clump on one member at a time, so the performance will be similar.

If you don't need to expand them, I would go with striped metas.

i can only comment on what kind of pain in the behind it is to come in into an environment that was provisioned as if it were a Clariion and start implementing local and remote replication. Devices are built from different size meta members,  some are striped , some are concat ..nightmare .

I think SAN_AK is planning to implement SRDF migrations for the same VMAX array.So I agree with dynamox. Have standard size devices  and use less no of meta members if possible.