Cloud costs are unmanageable: It’s time we standardize billing

More companies are adopting a multicloud strategy, which means they need to compare the costs and commitments they take on from the three major providers and choose services. Except, that’s nearly impossible. Google, Amazon and Microsoft bill so differently that many companies cannot achieve the benefits of a multicloud approach. They simply don’t know which provider is best for their requirements and usage. 

Gartner has forecast that end-user spending on public cloud services will reach $482 billion this year, a remarkable amount for something so lacking in transparency. Investment firm Andreessen Horowitz (aka a16z) has bemoaned how cloud costs drive down the value of public software companies by hundreds of billions of dollars. And some tech companies are saving enormously by repatriating operations from the cloud. 

Billing comparisons are nearly impossible, cost attribution is elusive

Nobody is questioning the value of cloud services themselves, but everybody understands their billing methods are a nightmare to untangle. There is too much at stake, and the numbers are too big, for this to continue. Standardized billing across cloud providers is long overdue. Here’s why. 

Non-standardized billing creates three sets of problems. The first is managing different types of commitments across cloud providers where the terms and implementations vary so vastly. The second problem is tracking expenses with different savings attribution schemes and cost metric definitions such as net amortized, unblinded, etc. being used across providers. The third is the increasing use of multiple cloud platforms and managed services within them, each with its own tagging conventions. For many, it’s virtually impossible to attribute costs internally even when using a single cloud platform. 

The net result is that customers cannot make an apples-to-apples comparison across providers. To understand the scope and complexity of this issue, let’s compare the three major cloud service providers: Amazon Web Service (AWS), Microsoft Azure (Azure) and Google Cloud Platform (GCP). 

The Big 3: Mature billing or not, all are confusing

Of the three, AWS has the most mature billing model. Here we define maturity as the number of discounted commitments available to customers as alternatives to on-demand purchasing. In 2019, AWS introduced Savings Plans to give customers another discounted purchasing model outside of Reserved Instances. This maturity has also allowed for AWS to develop the most granular pricing options per SKU. Increased optionality helps in selecting the best commitments to cover your infrastructure. But with so many choices, customers face confusion. For example, there are numerous obsolete billing constructs like Convertible Reserved Instances available that customers can mistakenly purchase in place of more efficient alternatives. 

Relative to AWS, Azure is less mature in their billing model. But they are more forgiving on things like enabling resale by providing guaranteed resale with a 12% penalty fee. For AWS users, there is a chance they’re stuck with Reserved Instances they can’t sell and don’t need. They also offer the additional option of a deeply discounted five-year commitment for certain resources, providing a price point that can actually compete with owning your own server. The other providers’ have a maximum commitment of three years. 

GCP is also less mature than AWS but does provide two discounted purchasing options. Committed Use Discounts provide a discount in exchange for a one or three-year commitment, like RIs and Savings Plans. GCP also innovated on the discount model by creating Sustained Use Discounts, which automatically apply discounts when compute engine VMs are used for a significant portion of the month. The threshold for the discount varies by resource type.

The independent development of each provider’s billing model has resulted in differences in how things are priced. Each “primitive” or component such as a machine, a managed service (like Lambda or Dynamo), bandwidth and storage all have different base pricing models that can be further complicated by long-term commitment discounts as well as top-level enterprise discounts. 

The benefits of having access to a wider range of services and the ability to choose is negated when you cannot make a comparison across services and have any confidence that it’s accurate. That’s why standardized billing is important to nearly all cloud users. 

How to fix this: Develop an open billing standard

Our team is currently working with the finops foundation and cloud customers to develop an open billing standard that can be used to compare projects using different vendors. 

The first area to tackle is creating a common standard to define the parameters for usage-based pricing of different components. This way you are not faced with comparing services that are charged by the hour with those that are charged by the amount of usage. The next is developing a common language to characterize commitment discounts between vendors and the level of flexibility the discount allows. This helps customers weigh the tradeoffs in using discounts that require a longer period of commitment, or offer some degree of additional flexibility, especially in cases where there may be variable usage. 

Allowing for an apples-to-apples comparison of SKUs will help customers select the right services for their needs across vendors. Customers won’t feel limited to using the vendor they are most familiar with. They can also rest assured that they are investing in the right resources to run their business optimally.

Aran Khanna is the CEO of Archera.