With the recent release of ONTAP 9.2 NetApp launched Fabric Pools - a technology that enables tiering of inactive data from EBS gp2, io1 or st1 to S3 for a lower cost.
Let us take a scenario whereby you have decided that you wish to leverage AWS in a failover scenario for your applications. ONTAP Cloud enables a highly efficient replication method via snapmirror and also maintains efficiency savings and snapshots. Once replicated this data though would typically sit on EBS , comparing the costs of EBS with S3 we see a stark contrast.
*EBS GP2 = $0.116 per GB-month of provisioned storage *S3 Standard = $0.024 per GB (First 50TB) , $0.023 per GB (Next 450TB)
So S3 storage is nearly 5 times cheaper than EBS and whilst this is not the full story needless to say the cost difference is not small.
If we were to deploy ONTAP in AWS we could leverage FabricPool to act as a gateway to S3 and reduce the overall costs of storage by having that data tier automatically into S3 as per below.
If you need to access the data to invoke DR the data will be tiered back into EBS.
Typically speaking not all of your data will be hot so therefore you may wish to deploy 25% of the capacity in EBS which you require and use S3 for the bulk. The EBS tier can always be expanded on the fly should that tier become full.
Additionally you are able to use st1 (throughput optimised) as the replication tier and then change to gp2 non-disruptively when data needs to be accessed. This along with being able to schedule uptime of ONTAP Cloud to only be on when a replication is occurring can change the dynamics of your cost model significantly.
Below is a short video of Fabric Pools in action for this specific usecase.
Even though the data sits in S3 we can still access the data through the usual protocol (in this case SMB). On access data is moved back to EBS as it's accessed and subsequent reads of that data will then be from EBS.
This capability also works with On-Premises all SSD aggregates and supports both Amazon S3 and NetApp StorageGrid Webscale.
Fabric Pool currently supports the following 3 use cases:
- Tiering of inactive snapshots
- Tiering of secondary data copies
- Archiving of inactive volumes.