Recently added Google and Microsoft cloud cost control tools are critical for users who want to gain insight into how corporate cloud budgets are spent.
Google and Microsoft have followed AWS' lead with more granular cloud budgeting controls, as customers struggle to decipher their expenses on these platforms.
AWS' public cloud launched years ahead of the Microsoft and Google clouds. This gave AWS a big head start to add helpful features, while the others learned the basics of how to operate at scale. Some of these features included greater transparency into consumption and more diverse payment schemes.
Microsoft and Google cloud services are now nearly on par with AWS. But a lack of visibility into their cloud platforms -- granularity that AWS also lacked in its early days -- has frustrated some Microsoft Azure and Google Cloud Platform (GCP) users. Over the past six months, both companies have attempted to address these shortcomings, particularly in the area of cost controls.
Azure, GCP answer user calls for billing insights
Microsoft did so predominantly through its acquisition of Cloudyn last year, which is now folded into Azure and renamed Cost Management. The service provides reports on usage and charges, and it continues to add APIs to track various categories of expenditures, such as recent support for tag-based filtering for cloud budgeting and usage for those with the Enterprise Agreements subscription. Other tools forecast future expenses, and updated alerts and monitoring tools provide consistent metering and pay-as-you-go pricing.
It's been difficult to link costs to resources in Azure, so this is a welcome step, said Moody Saada, site reliability engineer at Agolo, a startup in New York that provides AI-based information summarizations. Only one person can access the costs in real time, so the only way for various departments to review usage is to forward the bill that comes at the end of the month.
That bill's detail is limited, too. For example, it might state that the company used a certain product for a set amount of hours, but it doesn't say which environment, so usage can't be narrowed to a certain resource group.
"I still think scoping the costs or categorizing costs is very shallow," Saada said. "What I want to see as billing manager is to say, 'This is the resource, this is the environment, and these machines are costing me this much.'"
Meanwhile, GCP customers can now use Cloud Billing reports -- currently in beta -- to group costs by project, product or SKU. Users can narrow that data to monthly or daily usage and track the impact of service credits on their bill. They can export that information to BigQuery for data visualization in Data Studio, and there's a Cloud Billing Catalog API for real-time list pricing and predictive bill estimates.
If I was consuming something from Microsoft [or] Google, I'd want to get a report of who was consuming what ... That's not a crazy request.
Owen Rogersanalyst, 451 Research
Previously, the average GCP user had no real way to see what had happened with their bill without writing some CICS software, said Lukas Karlsson, cloud architect at the Broad Institute of MIT and Harvard in Cambridge, Mass. The Broad Institute's billing is additionally complex, with nearly 200 funding sources, so it uses the reseller model on GCP to split its bill and pass through to corresponding development teams to divvy up expenses.
Karlsson said he receives hundreds of emails a week for the various billing accounts, but it can be impossible to manually track those expenditures and know if they're in line with what should be spent by a particular department. The cloud budgeting tool has been one of the institute's biggest feature requests. And developers can now be aware of their expenditures, so they aren't unexpectedly shut down when they exceed their grant funding. It will also hasten responses to misconfigurations or coding mistakes that can often lead to overspending.
Owen Rogers
Much of the enhanced visibility into Azure and GCP was pushed by enterprise adoption of public cloud, as CIOs want tighter controls on spending there. So, while these updates are certainly welcome additions for users, it's really table stakes at this point, said Owen Rogers, an analyst at 451 Research.
"If I was consuming something from Microsoft, I'd really want to know who is consuming what. And if I was on Google, I'd want to get a report of who was consuming what," he said. "That's not a crazy request."
Customers have become more sophisticated in their understanding of cloud consumption and increasingly want metadata to provide details on cost allocation and chargeback, said Erik Onnen, CTO at Cloudability, a cloud cost management provider in Portland, Ore. Large customers can have more than 100 accounts on a given cloud, some of which will share resources across accounts, which can make it harder to understand broader application costs.
Multi-cloud compounds cost concerns
Cloud budgeting challenges are only exacerbated when a corporation spreads workloads across multiple public clouds, as there are nuanced differences between each vendor's products, terminology and charging practices.
The Broad Institute uses multiple public clouds, including AWS, which still has the most sophisticated cost management capabilities. The research organization is agnostic about these platforms, and Karlsson acknowledged the previous limitations on GCP are the types of issues that would give him pause about putting workloads on a platform.
Many companies turn to third-party cost management providers to help navigate all the differences between the various clouds and to get more detailed analysis on expenses. As part of its Cloud Price Index, 451 Research tracks half a million line-item tools across the major cloud providers. Because of that complexity, cloud providers likely will add machine-learning-backed tools that recommend best practices and automatically execute ongoing cost automation, Rogers said.
"Imagine how difficult it would be to try to get those 500,000 things in the price-optimized order," he said. "Realistically, it's not a task that can be accomplished using manual approaches."
Owen Rogers
