RHODF (now IBM Fusion as a Service) provides SAN-like software-defined storage, that will later be "consumed" by users for e. g. PVCs.

In it's TCO, there are three main components relevant for Cost Management:

• a) Cost of the underlying cluster ("provider cluster") where ODF will be running. They will be running on the customer's AWS account and paid directly by the customer.
• b) Cost of the ODF subscription. It will come as a marketplace service cost to the customer's AWS account. This cost is variable and based on the number of vCPUs of the connected application clusters.
• c) Cost of the underlying AWS EBS storage. Data is triplicated so 1 TB of available space requires 3 x 1 TB = 3 TB raw EBS space.

ODF natively supports deduplication and overallocation, which means the actually available storage space is 1 TB yet the aggregated total reserved by customers through PVCs created is e. g. 8 TB. Because of this, it's not really possible to determine what the unallocated capacity is (if any) and determine a "price per GB".

The strategy one customer shared is:

1. Sum cost of (a), (b) and (c)
2. Calculate total requested space (sum of PVC sizes), no matter if the provider cluster can actually satisfy those requests
3. Divide the cost calculated in (1) by (2) to calculate "a cost per GB" and then use that number to charge each PVC.

Note this requires attributing the cost of an ODF provider cluster to a different cluster (the application cluster). The relationship could be many-to-many: one ODF cluster may be providing storage to N application clusters, and N ODF clusters may be providing storage to 1 application cluster. We should evaluate if we need COST-3212 (multi-cluster cost distribution) before we can do this.