If your organization is like most, cloud services represent a large and growing share of your budget. In some of the companies I consult with, the operational expense of public cloud exceeds the capital costs of on-premises IT purchases, or soon will.
Analysts agree that software as a service (SaaS) is the largest category, accounting for about half of the cloud market. But infrastructure as a service (IaaS) is the fastest growing, with a five-year CAGR of 32%, according to IDC’s Worldwide Semiannual Public Cloud Services Spending Guide.
In real dollars, Gartner says that IaaS spending will total $38.9 billion this year, up from $30.5 billion in 2018, on its way to $76.6 billion by 2022.
The writing on the on-premises data center wall is clear.
CFOs want to support growth and agility while managing cash flow and establishing reasonable controls and guidelines. But cloud expenses are notoriously hard to track, contain and allocate.
In my engagements I have found that four key questions deliver clarity into cloud spend and open a path forward for productive conversations between finance and IT teams. For this column, I’m focusing on IaaS, specifically public cloud services, as it’s the fastest growing area and offers more opportunities for savings versus SaaS or PaaS.
Most IaaS providers have adopted a per-hour, per-GB pricing model for CPU, RAM and storage. Bottom line: It’s essentially pay-as-you-grow, with little incentive to scale down — which accounts for the steep upward trajectory in IaaS provider revenue.
It’s up to IT to architect services with cost efficiency in mind. There are three areas to consider here: overprovisioning of resources, idle or orphaned resources, and mis-provisioned resources. Let’s address those in our first few questions.
Seasoned IT folks have traditionally, and understandably, engineered on-premises environments to accommodate peak usage and ensure systems are available 24/7. Applying that same logic to public cloud, the IT team wants to avoid resource shortages that can cause application performance issues and result in unhappy end users or customers.
As systems move to the cloud, I often see clients rely on application vendors’ recommendations regarding the cloud resources needed. The result is almost always an exact replication of on-premises systems, except in the cloud. This may well be over-provisioning, resulting in additional costs.
Flexera, a company that offers technology asset management software, performed 60 initial cloud cost assessments for customers before those companies started their optimization efforts. It found that, on average, 35% of spending is wasted. One culprit: 40% of virtual machine instances are at least one size larger than needed, meaning companies are paying 50% to 75% more than they should be.
There are various guides on cloud migration strategies. IT should not simply default to “lifting and shifting” on-premises systems, such that the application runs exactly the same on the cloud platform as on a physical, on-premises server. Replatforming (making minor changes to an application to take advantage of cloud) or fully refactoring (completely reworking or recoding to run natively on the cloud platform) will take time and money but can pay off in stability, better performance and uptime.
Cloud waste doesn’t happen only in production environments; that is, the systems on which the business runs on a daily basis. I have seen virtual machines used for development, testing, staging and quality assurance sitting idle. That’s a problem because the cloud cost clock is ticking 24/7. Most IaaS suppliers bill hourly. If we use a 31-day month, that’s 744 hours that your firm is getting billed for, whether the resource is in use or not.
Simply turning off the virtual machines used for test and development on weekends can shave 192 hours from your bill every month. Shutting down those resources outside of all working hours, not just weekends, saves even more. If you are paying to run cloud infrastructure outside of a developer’s work hours, you are paying 50% or 60% more than is necessary.
Solving for both oversize and idle VMs is possible. IT needs automated or manual policies to shut down workloads after hours, eliminate inactive storage and right-size instances. Cloud management platforms, such as CloudCheckr, Flexera and VMware CloudHealth, provide visibility and control over cloud computing costs, performance and security. These platforms often include predictive analytics tools that can forecast future spend, offer right-sizing insights and identify unused resources.
Ask IT whether they have a comprehensive view of all the IaaS resources you’re paying for. I’ve heard from clients, upon getting a single view in a dashboard, that they found virtual machines they didn’t even know were running.
The cost for platforms like CloudCheckr is generally a small percentage of your monthly cloud spend. In my experience, first time users of these tools see an average 10% to 15% savings right away.
A recent survey by 451 Research, sponsored by Cloudability, shows 82% of respondents rely on spreadsheets and manual tracking or cloud vendor portals — or have no visibility into cloud costs at all.
I guarantee that these companies are paying significantly more than they need to.
For example, if you use Azure and commit to one- or three-year terms, you can save up to 72% over pay-as-you-go pricing. That’s best case, but you can realize savings even without signing up for the long haul. Microsoft EAs offer discounts from 15% to 45% percent for Azure users, depending on the level of commitment. Additional discounting can come from the use of reserved instances, Hybrid Use Rights and the ability to transfer owned SQL licensing.
AWS has similar savings, and Google’s sustained use discounts (SUDs) happen automatically and require no upfront commitment. Google charges according to the percentage of usage in a month.
Yet RightScale’s 2019 State of the Cloud report shows that companies are not taking advantage of discounting. Among AWS users, only 47% use reserved instances, while Azure users leverage reserved instances only 23% of the time.
While pay-as-you-go is the default method and is highly flexible, it is not cost-effective.
Every cloud provider offers pricing structure flexibility and discounting, and most offer tutorials on optimizing costs. You don’t need to make long-term commitments to save, either.
I recommend asking IT to pull together discounting options for all the IaaS suppliers you purchase from, usage history reports and growth projections as far out as is feasible. Gather detailed cost reporting and analysis to baseline your cloud spend. Then make contractual decisions based on data. It takes more time, but the savings are well worthwhile.
A cloud management platform can help here as well. These systems include advanced cost optimization tools and can not only spot areas where reserved instances are a fit, they can make recommendations on rebalancing workloads to cover for unallocated reservations that you may have already purchased.
In addition, turn on alerting and monitoring at both the cloud platform and application level. This helps keep cost thresholds from being overrun.
There is a truism in IT that user demand will expand to consume all available resources — bandwidth, storage, CPUs, whatever. A best practice to check that urge is using chargebacks to make lines of business aware of their resource use. In my practice, clients that haven’t done this see over-provisioning abuses because application owners aren’t incentivized to dial down unneeded infrastructure. You’ll also need to create tagging standards and ensure all resources are properly sorted for cost allocation.
Organizations I work with that have mature cloud practices have learned that dashboards and reports that provide consumption insights on various groups, from developers to executives, make a real difference.
In fairness, your IT department may not have visibility into all the cloud services being used by departments and lines of business — when any employee can use a credit card to buy some VMs, spending becomes the Wild West. There are tools that can rein in shadow IT, but they can be expensive and complex. Visibility is key, and CFOs have insight into where money goes. You can help the CIO catch not just wasteful cloud spending but potential compliance and security risks.
Cloud is a journey, not a destination, and until the IT team gains some experience using public cloud, right-sizing your environment is a mixture of art and science. Philosophy aside, cloud optimization is not a one-and-done process. Once you understand spend volume, continue to engage with key stakeholders to drive and maintain accountability. Put mechanisms in place that allow you and your team to check in regularly. Plan on quarterly touch points. Ensure that there is an actionable, automated plan to eliminate wasted cloud spending.
Jo Peterson is the vice president of cloud and security services for Clarify360. Her engineering team focuses on cloud enablement and cloud security. As an engineer and leading industry expert, Jo sources net new technology footprints, as well as optimizing and benchmarking existing environments.
Jo is currently ranked as #5 on the Rise Global Cloud Power 100 as well as #2 on the Rise Global Cybersecurity Power 100, a Top Woman in CyberSecurity by CyberCrime Magazine and a 2019 Channel Influencer by Informa. She is a member of the Forbes Technology Council. Followed by an audience of over 45,000 on Twitter, Jo is a featured keynote speaker on topics of cloud architecture, procurement and management, and cloud security. Jo has provided social media thought leadership and independent analyst insight to brands such as Intel, HPE, IBM and others at events such as vmWorld, Think, Discover, AWS reinvent, Google Next and many others.