Fabric pool design

[Jackiez]

Hi, Experts,

       One customer will have 1PB hot data and 10PB cold data(from one application) to do data tiering. Since he has two Datacenters (DC-A and DC-B ,  several KMs far away, DC-A is primary, DB-B is secondary), and since DA-A has not enough rack space , so has to split cold data storage into DC-A and DC-B.   My question is if we design a Fabric pool solution like below , is it workable ?  And can i manage which data to DC-A or DC-B cold tier ?  How to improve read performance if need read cold data from DC-B ?

      Draft design as this :   Put one SSD storage (1PB) in DC-A as hot data storage , put one SG or ONTAP S3 storage (5PB) in DC-A as part of cold tier, put another SG or ONTAP S3 storage(another 5PB) in DC-B as another part of cold  tier .  

     If this design is not workable ? Any other suggestions ?

     Rgds,

     Jackie

[AlexDawson]

FabricPool is attached to an aggregate. All tiering for an aggregate is done to a single S3 destination (migration is supported). At 6Pb, you significantly exceed maximum aggregate size, so you'll need to have several buckets. Splitting buckets between DCs means you are reducing cross site isolation, reducing resiliency.

At this sort of scale, NetApp will be happy to provide you architectural resources to assist - have you been in touch with the account team for their support?

Hope this helps!