Why managing your cloud and SaaS spending is trickier than you might think
Inefficient software, over-cautious managers and buyers with little oversight can all make the cloud more expensive than anticipated.
One of the main reasons companies cite when moving to the cloud is the money they'll save now that they no longer have to pay somebody to go find that misfiring hard drive somewhere in a cavernous data center. The reality is more complicated.
While the cloud can help companies save money and definitely buys convenience, converts often encounter sticker shock when they make the switch. Done properly, moving to the cloud actually increases short-term costs given all the work needed to migrate critical applications without downtime. Done improperly, moving to the cloud can dramatically increase overall costs as applications designed around the assumption of unlimited hardware resources now run on a meter.
And when developers have nearly unfettered access to computing resources and software-as-a-service is bought on a company credit card, the ease with which projects can be started and abandoned, only to rack up charges for months on end, can drive the CFO crazy.
"Some businesses focus maybe too much on their topline revenues, and focus all their efforts on increasing topline revenue numbers, but saving a dollar on your Amazon spend, that's directly affecting your bottom line," said Pete Cheslock, cloud economist at The Duckbill Group, a consulting firm that helps companies manage cloud spending.
The good news is that there are a lot of software tools that can help cloud customers get a handle on their spending, and cloud providers have also acknowledged the need to get better at helping customers understand how and where they're spending money. A trend toward multiyear contracts with fixed amounts of spending each year also allows the finance folks to plan more confidently for the future.
Yet, like everything else it has touched this year, the pandemic upended conventional thinking about the economics of the cloud.
Those multiyear deals that looked like prescient cost-saving steps became albatrosses around the necks of companies in hard-hit industries such as Airbnb, though it was able to renegotiate its deal with AWS to stay on track towards its imminent IPO. Scaling down is trickier than scaling up if you haven't taken steps to redesign your application around some of the more nimble cloud technologies, such as containers or serverless.
Still, there's no doubt that the cloud was a lifeline for businesses forced to rethink elemental aspects of their strategy on the fly during the first half of the year. As those businesses settle into life on the cloud, though, they'll need to find long-term ways to manage this new way of doing business.
Searching for the 'cruft'
Most cloud customers probably aren't spending enough with their provider to be eligible for volume discounts, said Eugene Khvostov, vice president of product and engineering for Apptio, which makes tools that help companies track and manage their technology budgets.
At the same time, they're being asked to do more with less. According to Apptio's data, "over 60% of companies reported there's an increase in demand for new technology capabilities that they started their business with, but no change [to] — or in some cases, even reduced — budgets for [those] same technologies," he said.
So how can they save money? In lots of cases, cloud customers can find a fair amount of "cruft," as Khvostov put it, simply by auditing their own usage of cloud services.
One common example is storage. AWS offers different types of storage for data on its S3 platform: Some of it, for information that's needed on a less frequent basis, is very cheap to store but expensive to access. Data placed in a bucket designed for that sort of long-term storage can quickly rack up costs if an application is constantly requesting that data. Some of the storage, meanwhile, is designed to allow frequent access to stored materials at a cheap rate, even if the storage cost is a little higher.
Other savings can often be found in applications that have been "lifted and shifted," or basically moved from data centers to cloud servers without any architectural changes. Cloud providers offer an increasing array of managed services using containers and serverless technologies that can allow customers to spin up or cut off computing resources in seconds or less. That can eliminate a lot of waste compared to the virtual machines, typically rented by the hour, that a lot of ported software is likely still using.
"A lot of cost savings do take some engineering work to do," Cheslock said. Companies disproportionately impacted by the pandemic are also finding this out the hard way, especially if they optimized their cloud strategy around growth.
There are two main reasons why companies over-provision cloud resources: They're worried about providing a subpar experience to their users, or they're worried about having to page an engineer every time they hit a new capacity limit. When their businesses are growing, companies are happy to spend a little extra for peace of mind. But when demand falls off a cliff or shifts to an entirely different part of an application, pulling back can be harder than it might seem — especially if a company has purchased reserve instances, which are generally cheaper but locked in.
Other companies are reluctant to use anything beyond what Cheslock called the "core primitives" — the basic compute and storage services that all cloud providers offer — out of fear they'll be locked into higher-level services. But platform-as-a-service offerings such as AWS's Fargate or Microsoft's Azure Container Instances often promise cost savings, or at least a great deal more operational flexibility, he said.
"Because of that [reluctance], there's a lot more waste inherent in their spending," Cheslock said.
Who has the credit card?
Companies are also realizing that they need to get a better handle on their spending with software-as-a-service vendors, who have come to dominate the market for the software needed to run a modern business.
"There's generally a lack of centralized oversight and accountability," Khvostov said. Most enterprise SaaS companies make it very easy to get started with free trials or low-cost tiers for small teams, which can then ramp up very quickly over time, he said.
And usage of enterprise SaaS services has exploded over the last several years.
"The average customer we see in the enterprise segment has over 200 applications," Khvostov said. Increasingly, those applications also charge on a consumption basis rather than a subscription basis. "That's only going to accelerate the need to be more conscious, more centrally aware, before we get surprised by it," Khvostov added.
Ideally, getting a handle on this spending often comes down to encouraging people to better manage their budgets. AWS has added some cost-management services over the last year or so, but actually charges customers for those services, while Microsoft and Google offer cost-management tools as part of the basic subscription.
Businesses are increasingly investing in tools that can help those responsible for purchasing tools for their department — a task that has become much more decentralized over the past few years — understand how much of their budget they are spending, and work to set parameters around that spending, Khovstov said. If that doesn't work, they can also set the finance department's version of parental controls: Spend too much in a given quarter, and funding is cut off.
There's no doubt that for lots of companies, cloud computing can be a much more efficient way to consume computing resources than managing data center hardware and large packages of enterprise software. However, like many things in life, realizing those cost savings is not necessarily as simple as it might sound.
"The problem is the people side of things; it's the people underneath that are making choices that cause the numbers to change, and prioritizing internally on what you care most about is a challenge," Cheslock said. "In the classic data center space, you spent the money; the money is spent, the servers are there, and you don't have to spend any time optimizing spend."