...with introduction of native ZFS support and HDD /SSD sizes (cost/TB) still decreasing, the concept of using the array as "cold" data store and pools as "hot+fast" data store gets more and more appealing.

 

Using shares on a pool is only possible with setting the named pool as a cache pool for these particular shares.

However, with the concept of a "hot+fast" data store in mind, the actual caching concept is not applicable to pools themselves.

 

Proposition:

 

• introduce a concept to distinguish pools into type "data" and type "cache" types of pools.
• add a feature to allow shares to be defined on pools of type "data"
• allow pools of type "cache" as cache for shares, regardless of where these are declared for (array or type "data" pools)

 

...hence allowing cascading pools (a data type pool can have a cache pool, similar to the array).

30 minutes ago, Ford Prefect said:

...with introduction of native ZFS support and HDD /SSD sizes (cost/TB) still decreasing, the concept of using the array as "cold" data store and pools as "hot+fast" data store gets more and more appealing.

 

Using shares on a pool is only possible with setting the named pool as a cache pool for these particular shares.

However, with the concept of a "hot+fast" data store in mind, the actual caching concept is not applicable to pools themselves.

 

Proposition:

 

• introduce a concept to distinguish pools into type "data" and type "cache" types of pools.
• add a feature to allow shares to be defined on pools of type "data"
• allow pools of type "cache" as cache for shares, regardless of where these are declared for (array or type "data" pools)

 

...hence allowing cascading pools (a data type pool can have a cache pool, similar to the array).

I think it is planned that you will get that (in 6.13 or later) when the distinction between the current main array and pools is meant to be removed (the current array type will become another pool type).