Cloud Hosting Blended Cost for Azure China

Objective

If we were to explain on a user story format what we want for this article, it would be:

As a user, I want a new type of cost that matches the net total on azure invoices and the commitment based discounts are fairly distributed throughout the teams.

Some concepts here were covered on my other blog post about Cloud Hosting Cost Types. As Azure China behaves differently from Azure Global, this article aims to provide the business requirement to have the same concept of Amortised and Blended cost we have on Azure Global but now applied to Azure China.

Background

Commitment based discount in Azure China is called Compute Pre-Purchase (CPP). Differently from Reserved Instance (RI) from Azure Global, the CPPs are acquired and applied in a yearly and/or monthly basis. It is attractive from an economic standpoint, as it reduces the chances of having unused CPPs. But on the other side, CPPs bring a challenge for cost reporting as the cost of its acquisition and usage is not included in the billing events provided by Azure China.

It was verified with Microsoft and, although there are plans to migrate Azure China into Azure Global model and replace the CPPs by the Reserved Instances, this is unlikely to happen until mid-2023.

In order to provide cost reporting as close as possible to the real (invoiced) cost, it is necessary to add logic to Hosting Cost Reporting so that users can have access to blended and amortised cost the same way they have on Azure Global.

Blended Cost Calculation for China

Here is an example that explains how blended cost could be calculated for Azure China hosting cost data. From now one, for simplicity, every time blended cost is mentioned, it is related to China.

Blended cost should be understood in four moments in time. Let’s use the month of June of any given year as a reference.

IMPORTANT: This logic will also be applied to a grouping of SKU and contract, as the CPP can vary according to the contract in Azure China.

Moment 1: Before the Month (M-1)

Each month the FinOps team decides** the quantities of each SKU will be subject to CPP on the following month. This decision is logged in a separate spreadsheet

** the process on how to calculate the SKU quantities are out of the scope of this requirement. Simply consider the data on the table as a given. We are also simplifying the explanation by assuming that we have monthly CPPs only (no yearly CPPs).

Moment 2: During the Month (M-0)

During the month of June, consider that just two SKU were used, and all billing events are the ones on the table below. SKU A is subject to CPP price while SKU B will incur a pay-as-you-go price or full price.

As the billing events are received, the Hosting Cost Report uses the price of the CPP Spreadsheet for SKU A to calculate the cost by multiplying it by the Quantity. See that SKU A will use the CPP Price while SKU B will use the pay-as-you-go or Full price.

Moment 3: After the Month (M+1) — Price Recalculation

At this moment, when the month of June is closed, it is time to reconcile the real price for SKU A. The actual resources consumption will naturally be different from the forecast considered by the FinOps team at Moment 1. In this example, the actual quantity is higher than the quantity planned on Moment 1. In this case, from the total of 790 actual units, 744 units are charged using CPP price while the other 46 units are charged using the Full Price.

Moment 4: After the Month (M+1) — Cost Recalculation

Now, with the average price (or blendedprice) calculated, all costs related to the events from the month of June are recalculated. This will impact the cost of SKU A while no change is expected on SKU B.

Implications

• Differences between the CPP quantity and Actual quantity for a given SKU on a given month will result in distribution of these differences among all billing events of this SKU. This means that the cost of a team in that given month can change after this given month ends.
  – If the difference is low, most likely the impact will not be significant.
  – If the difference is high, it is possible that this can impact cost significantly.

Benefits

• Teams will be able to have blended costs for Azure China at every billing event, as soon as they are made available by this Cloud Provider.
• Consider that FinOps Team history tracking on CPP quantity against actuals is accurate. In this case, cost recalculation at M+1 should not cause a significant impact on M-0 month.

Conclusion

One of the FinOps principles says that reports should be accessible and timely. Azure China’s Compute Pre-Purchase (CPP) provides great savings, but doesn’t allow cost visibility before the month end. The Blended Cost Calculation here described help with two challenges: Have a timely cost reporting and produce a blended cost that is balanced according to resource usage by teams and services.

References:

CPP (Azure China Compute Pre-Purchase) vs RI (Azure Reserved Instances) — cloudeasier.com

Purchase of Azure China | Microsoft Docs

CPP (Compute Pre-Purchase) | Azure Docs

Originally published at https://www.farfetchtechblog.com on December 19, 2022.