We’re excited to bring Transform 2022 back in person on July 19 and pretty much July 20-28. Join AI and data leaders for insightful conversations and exciting networking opportunities. Register today!
More companies are adopting a multicloud strategy, which means comparing the costs and obligations they incur from the three major providers and choosing services. Except, that’s almost impossible† Google, Amazon and Microsoft bill so differently that many companies cannot realize the benefits of a multicloud approach. They just don’t know which provider is best for their needs and usage.
Gartner has predicted that end-user spending on public cloud services will reach $482 billion this year, a remarkable amount for something that has so little transparency. Investment firm Andreessen Horowitz (aka a16z) has complained that cloud costs are reducing the value of public software companies by hundreds of billions of dollars. And some tech companies are making huge savings by repatriating operations from the cloud.
Billing comparisons are almost impossible, cost allocation is elusive
No one is questioning the value of cloud services themselves, but everyone understands that their billing methods are a nightmare to untangle. Too much is at stake and the numbers are too great for this to continue. Standardized billing between cloud providers was long overdue. This is why.
Non-standardized billing poses three problems. The first is managing different types of obligations with cloud providers, where terms and implementations differ so wildly. The second problem is tracking expenses with different savings attribution schemes and cost metric definitions such as net amortized, unblinded, etc. used by 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 is virtually impossible to allocate costs internally, even when using a single cloud platform.
The net result is that customers cannot make an apple-to-apples comparison between providers. To understand the magnitude and complexity of this problem, let’s compare the three major cloud service providers: Amazon Web Service (AWS), Microsoft Azure (Azure), and Google Cloud Platform (GCP).
The Big 3: Legacy Billing or Not, They’re All Confusing
Out of the three, AWS has the most mature billing model. Here we define maturity as the number of discount commitments available to customers as an alternative to on-demand purchases. In 2019, AWS introduced Savings Plans to offer customers another low-cost purchasing model outside of Reserved Instances. This maturity has also allowed AWS to develop the most detailed pricing options per SKU. Increased optionality helps select the best pledges to cover your infrastructure. But with so many choices, customers are faced with confusion. For example, there are plenty of outdated billing constructs like Convertible Reserved Instances available that customers may accidentally buy instead of more efficient alternatives.
Compared to AWS, Azure is less mature in their billing model. But they are more forgiving about things like enabling resale by offering guaranteed resale with a 12% penalty. For AWS users, there’s a chance they’ll be stuck with reserved instances that they can’t sell and don’t need. They also offer the added option of a heavily discounted five-year commitment on select resources, delivering 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 it does offer two low-cost purchase options. Committed use discounts offer a discount in exchange for a one- or three-year commitment, such as RIs and savings plans. GCP also innovated on the discount model by creating extended use discounts, which automatically apply discounts when compute engine VMs are used for a significant portion of the month. The discount threshold varies by resource type.
The independent development of each provider’s billing model has led to differences in the way things are priced. Each “primitive” or component such as a machine, a managed service (such as Lambda or Dynamo), bandwidth, and storage all have different base pricing models that can be further complicated by long-term discounts and top-level corporate discounts.
The benefits of access to a wider range of services and the ability to choose are negated if you cannot compare services and be sure it is correct. That’s why standardized billing is important for almost all cloud users.
How to fix it: Develop an open billing standard
Our team is currently working with the finops Foundation and cloud clients to develop an open billing standard that can be used to compare projects with different suppliers.
The first area to address is to create a common standard to define the parameters for usage-based pricing of different components. This way you are not confronted with comparing services that are charged per hour with services that are charged per usage amount. Next is developing a common language to characterize commitment discounts between suppliers and the degree of flexibility the discount allows. This helps customers weigh the trade-offs when using discounts that require a longer commitment, or offer some degree of added flexibility, especially in cases where there may be variable usage.
By allowing an apple-to-apple comparison of SKUs, customers can select the right services for their needs from different vendors. Customers will not feel limited to using the supplier 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†
Welcome to the VentureBeat Community!
DataDecisionMakers is where experts, including the technical people who do data work, can share data-related insights and innovation.
If you want to read about the very latest ideas and up-to-date information, best practices and the future of data and data technology, join us at DataDecisionMakers.
You might even consider contributing an article yourself!
Read more from DataDecisionMakers