Expansion Shelf In Scale Out Design

[crw64]

I understand that expansion shelves are treated as their RAID group. I am trying to understand how this plays into an environment where multiple storage controllers may be combined in the same storage group where data is being striped across multiple storage arrays. Ie. 1 x CS460x2 + 1 HS65, being combined with a second CS460x2 My assumption is that the expansion shelf would be recognized and managed as a separate storage pool in the same storage group and would be independent of the 'combined' storage pool shared between multiple arrays. Is this correct?

I have seen the documentation around scale-up and then scale-out, but none of these documents takes into account a single array with multiple expansion shelves, being combined with a second, third or fourth storage array with no expansion shelves.

Thanks

[Nick_Dyer · Accepted solution]

Hi Corey,

In the world of Nimble an expansion shelf is indeed it's own RAID set, however it is owned and managed by the pair of controllers in the head shelf, and the capacity of both the head shelf and expansion shelf are combined together with the controllers running a traffic policer to determine how the data can be efficiently distributed across both shelves of available capacity.

When scaling out with multiple systems in a group, an expansion shelf is seen as part of one of the arrays in the group. It's possible to create multiple pools within the group (see Ezat's excellent Pools blog post for more information) however a pool encompasses one or more systems which may or may not have expansion shelves on them, rather than creating a pool for a particular expansion shelf on a particular system.

Hope the above helps!

[crw64]

Thanks Nick. That does make sense.

Thank you.