Denys Rudyi - Fotolia
As the COVID-19 pandemic and economic factors drive more workloads to the cloud, enterprises that don't adopt robust cost optimization practices do so at their peril.
"I think we're in a perfect storm of factors that are going to exacerbate complexity and make cloud cost optimization more important than ever," said Owen Rogers, an analyst with 451 Research, part of S&P Global Market Intelligence.
The pandemic forced enterprises to prioritize implementations at the expense of optimization and management concerns, Rogers and fellow 451 Research analyst Jean Atelsek wrote in a recent report. Steep drops in compute costs thanks to improved processors won't have much benefit to many companies, however, since their usage went up.
"This is a vicious cycle," they wrote. "Deployment of more resources and services leads to increased complexity and higher costs. The longer it takes to optimize this complexity to deliver business value, the more complex the IT estate becomes."
That said, many enterprises are making cloud cost optimization a priority in 2021 and beyond.
Change Healthcare, a $3.3 billion health IT company in Nashville, Tenn., has developed a comprehensive FinOps practice in recent years, said Tatum Tummins, R&D effectiveness manager, during VMware's CloudLive 2021 virtual event last week.
Owen RogersAnalyst, 451 Research
Previously, Change Healthcare depended on its operations and engineering teams to manage all facets of the cloud, including spend, he added. Some 70% of cloud costs had no clear owner within the company, with the majority of accounts shared by disparate product teams.
Change Healthcare, like other companies that presented at the conference, uses VMware's CloudHealth software to track and optimize cloud costs by tagging resources used on AWS and other clouds. The software can be used to chargeback and showback cloud spending to various teams and departments, optimize spending with models and recommendations, and add a governance layer to provide alerts for spending anomalies or potential budget overruns.
It's critical to give access to tools like CloudHealth to a broad swath of employees, from application and operations teams to the finance department and line-of-business leaders, Tummins said.
"You've now enabled all of your stakeholders to participate in discussion about costs related to the cloud," he said. "I can't tell you how often this is overlooked. We ran into situations where people didn't know they were allowed to use CloudHealth or how to get access."
To gain adoption, tools like CloudHealth need an "ambassador" within the company who can evangelize its benefits and provide basic training, Tummins said.
"You're going to have some teams that are early adopters … but you're also going to have teams that don't see the value right away," he said.
It's also important to understand each product team's environment and needs. "For some teams that don't have a lot of new enhancements or development going on, they might have 24/7 workloads that don't change," he said. "One size doesn't fit all for all of your teams. You need to have context before you can make decisions about how to better save money for your company.
FinOps, meet SRE
SaaS business intelligence software vendor Domo, Inc. has worked to align its site reliability engineering (SRE) teams with a FinOps initiative, said Riley Jenkins, senior SRE/FinOps architect, in another CloudLive presentation.
"When you're doing SRE, a lot of it comes down to making [the application] faster," Jenkins said. That can put questions of infrastructure cost at a lower priority. But Domo has found that a holistic approach pays dividends.
Domo's SRE team had a strong relationship with other parts of the business, which helped the move into FinOps. "We already had a lot of trust from our engineering teams, so when we started saying you could save a lot of money doing XYZ, they listened," Jenkins said.
One big effort centered on developing a tagging and cost identification system. "That's the biggest amount of work, understanding the business context, and not just the engineering context," he said. "You have to implement a tagging policy that works for both worlds."
Beyond that, Domo keeps tabs on new or different AWS services that could bring a better payoff.
For example, the company built a tool that allowed it to use AWS Spot Instances effectively in production, which can be a tricky proposition.
Spot Instances use idle capacity on AWS servers and can cost as much as 90% less than on-demand instances. Customers bid on the price they're willing to pay, but the instances get turned off, or "evicted," once the market price rises above the customer's chosen cost. This can potentially lead to interruptions or slowdowns in an application's performance.
In September, the AWS Spot Instance market began to shift in a direction Domo didn't like, with costs and evictions on the increase. "We started seeing some rocky times."
Domo investigated the use of AWS instances based on the cloud vendor's homegrown Graviton 2 processors. Its analysis showed that Graviton could be cheaper or the same as Spot Instances when purchased on a reservation basis.
Domo took a system that ran almost completely on Spot Instances, moved it to a group of large Graviton instances, and found improvements in performance, scalability and cost, according to Jenkins.
Decentralization's cloud challenges
Robert Half International, a $5 billion staffing firm, has used AWS since 2010. The company operates in a decentralized manner, which has resulted in more than 200 AWS accounts owned by individual business units, said James Fogerson, senior manager of the cloud center of excellence.
Accounts are requested through ServiceNow, and subsequently created under a master management account, which allows for consolidated billing.
"[Account holders] can do whatever they want based on the needs of their business," although AWS Reserved Instances are largely purchased centrally in the master account, Fogerson said. Last year, Robert Half began investing in AWS Savings Plan, another discounting program.
Savings Plans have dollar-per-hour pricing, are available in every region, and apply any instance family, any tenancy and any OS. "It's a lot more flexibility," he said.
In contrast, Reserved Instances are sold on a per-instance basis and are specific to AWS regions and zones. They can be converted as technical needs change, but this requires some effort. Another key difference is that while Reserved Instances provide discounts off list price based on a customer's committed usage, Savings Plans' discounts are based on a customer's spending commitment.
"The discount seems to be the same but there is the added benefit of increased flexibility [with Savings Plans]," Fogerson said. "I no longer needed to spend lots of time converting instances anymore."
There's still the matter of tracking AWS spending across business units and communicating when there's an anomaly or an opportunity to tweak usage and save Robert Half money.
Robert Half's cloud center of excellence uses CloudHealth to generate daily, weekly and monthly reports that break down AWS spending. CloudHealth also offers benchmarking KPIs that lets Robert Half compare its spending against other companies.
The company's decentralized approach makes cloud cost management more time-consuming and complicated, but it's worthwhile, Fogerson said.
"I think a lot of people, at least when we started [in the cloud] 10 years ago, had the perspective of everything had to go through IT and it slowed things down," he said. "One of the more important things we've done in the cloud is giving the business opportunities to investigate new technologies."