The objective of this change is to ignore, for purposes of HCS customer’s drawdown calculation, any entries in the customer’s Cost and Usage Report (CUR) where lineitemtype is equal to RIFee.

For AWS the CUR contains two sets of entries for each usage of a Reserved Instance (RI). The first set includes the upfront payment, if any, the customer pays for the RI and the amortized cost of the monthly charges for the RI. The sum of these over the period of the RI equals the total amount paid for the RI. Unfortunately, the upfront fees are in CUR entries where the lineitemtype is Fees and have a generic lineitemdescription so it is impossible to determine which of the RI charges are for instances where Red Hat products are included.

The second set of entries is the usage of the RI, which shows the customer additional useful information. For CSI’s purpose that information includes the number of usage hours, a lineitemdescription that shows Red Hat products are included in the instance and the number of vCPUs in that instance. These lines have a lineitemtype of “DiscountedUsage”. Summing the estimates of cost for these entries would show the total amount paid for the Red Hat portion of the RI.

To implement this change and meet the needs of CSI, at least two options exist:

1. The filtering for Red Hat CCSP spend could ignore lines where lineitemtype is RIFee. Something like:

• line_item_legal_entity includes “Amazon Web Services”  AND line_item_line_item_description includes “Red Hat” AND lineitemtype is not “RIFee”
• line_item_legal_entity includes “Amazon Web Services”  AND line_item_line_item_description includes “RHEL” AND lineitemtype is not “RIFee”
• line_item_legal_entity includes “AWS”  AND line_item_line_item_description includes “Red Hat” AND lineitemtype is not “RIFee”
• line_item_legal_entity includes “AWS”  AND line_item_line_item_description includes “RHEL” AND lineitemtype is not “RIFee”

1. Or, using the existing filtering, change the spend estimate (unblendedcost) to zero for lines where lineitemtype==RIFee.

The filtering option seems to be the cleanest approach, but both are accurate.