Is Cloud Cost Optimization the Next Fad Diet?
By Etienne Gadient
Exploring how many enterprises look to put their cloud spend on a diet after adopting the public cloud and finds that many fall into a common trap as they look to build and execute a Cloud Cost Optimization initiative.
All I need is a tool right?
Today, clients tend to focus on Cloud Cost Optimization through one or more tools, usually a service supported by a tool (or tools) analyzes one or more cloud tenants. Through the analysis, they find mis-sized and mis-configured elements (typically compute), along with other areas of waste and suggest corrective actions that will directly reduce monthly cloud billings.
Cloud Cost Optimization comes in several varieties and focus areas, each with unique pros and cons. For example, the market has several Cloud Cost Optimization tools developed by both the Cloud Provider (Amazon Trusted Advisor, Microsoft Cloudyn) and third party cloud management tools (CloudGenera, CloudHealth, CloudCheckr, HyperGrid, etc.).
The irony about these different approaches is that clients see Cloud Cost Optimization as someone else's job. While it can be much more fun and exciting to talk about refactoring an application using the latest "sexy" platform tools, the reality is that there is much to achieve from a good cost optimization program. Right Scale, in its “State of the Cloud” 2018 report routinely finds 40%-50% cloud expense waste in surveyed companies. In 2019, that number is still over 35%.
Let me highlight that a different way.
On average, for every $1 that is spent on cloud, over 1/3 is thrown away… No, that is not a joke, and to our experience, it is not wrong either. The irony is that clients routinely underestimate their waste...who routinely survey their waste at closer to 15-25%.
Causes of Waste in Cloud
Research finds several major causes that led to waste in cloud deployments:
1. Unused Virtual Machines (VM) and Storage
2. Running all VMs 24x7
3. Oversizing Resources (storage, memory, instance sizing)
4. Disastrous Financial Choices
- Not leveraging Reserved or Spot Instances where possible
- Lack of storage lifecycle policies
- Failure to continue innovating the application set
5. Sub-Optimal Application Placement
6. Focusing on optimizing compute while ignoring storage and network hidden costs.
The good news is that most of the tools above will help you find most of these 5 areas.
The bad news is that the enterprise has to be judicious and committed to making the changes. At the end of the day, outside of hiring a managed service or professional service provider that actually executes the changes to optimize (typically costing you an additional 15% of your cloud spend…), a report can only help you identify what to do…not do it for you.
Finding a home for Cloud Cost Optimization to thrive
Within an enterprise, where should the responsibility for Cloud Cost Optimization live? This is where the debate gets very interesting to our crew. On one side of the debate, should Cloud Cost Optimization be the responsibility of the budget holder? After all, it is their money being wasted…in theory, they will have the biggest motivation to aggressively drive out cost. Or should it live within IT as a shared service? This could be an important benefit of IT as a shared service where the best practices can be applied at a macro level. Or another approach would be to make it part of audit and finance, since they ultimately influence the size of next year’s budget?
We have found that there commonly is a better option for enterprises adopting cloud formally. In our experience, a recipe to successfully deliver Cloud Cost Optimization is the Cloud Center of Excellence (CCoE), specifically as a named and measured role. According to Amazon, a “Cloud Center of Excellence (CCoE) is a cross-functional team of people responsible for developing and managing the cloud strategy, governance, and best practices that the rest of the organization can leverage to transform the business using the cloud.” The CCoE should provide the reporting, insights and recommendations, in addition to having a committed team of enabled doers who absolutely hate giving Amazon, Azure, Oracle or SoftLayer a single unnecessary penny.
It gets better…By charging the CCoE with the responsibility of providing Cloud Cost Optimization as a service to the business means that the CCoE just went from a cost center to a profit center in the eyes of the executive leadership...Measuring every “Benjamin” saved and telling the story up high...After all, who doesn't love people who create free money (through savings)?
Locating Cloud Cost Optimization in the CCoE probably seems counter-intuitive. After all, cloud providers and many organizations who promote a CCoE see it as body focused on cloud adoption and governance. Yet, Cloud Cost Optimization benefits from being in the CCoE in many ways.
Why upside down is the new right side up
In traditional IT, planning and design is typically top down, with initiatives planned years in advance with a growth factor (usually 10-15% per annum), forecasted for anticipated target scale and a risk factor (padding). In other words, Infrastructure architects (and I am one of them) are taught to focus on risk first because it is typically harder to grow incrementally in an enterprise data center. As a result, we over-design from the beginning and let the organization grow into the capability…besides capital expenses are typically amortized over 3-5 years, so an extra 20% can be spread across that period is a minor overall impact. In other words, it is effectively cheaper to add a little waste as a buffer, than to fall short and fail disastrously.
In the cloud, experimentation and scale are some of the biggest advantages to leverage. Using “fail fast, fail often” as a mantra, design small and grow incrementally , as bottlenecks occur. Since consumption costs grow linear in scale to consumption, you can scale both horizontally (add additional nodes) and vertically (make each node larger) with the costs following the growth.
Migrate Safely, then Resize
Whether refactoring or migrating to infrastructure as a service, we've found it is actually better to oversize your initial workloads as they transition from your data center and then resize them. Don't lose the time focusing on getting the sizing right in the beginning…you won't…just make sure it is large enough to handle the transitional requirements and the operational load.
Once the transition is complete, baseline the workloads, perform cloud cost optimization and apply tools to guide you to ruthlessly right size and right scale the workloads.
Load First, Shoot Second, Aim Third
First, prepare for the impacts of the change. For more on preparation, fellow ATC member Matt Brown has an excellent article on “Why context matters for applications”. Once you have discovered, blueprinted, and packaged, migrate first and optimize second, third, fourth, and fifth. Oh yeah, and then do that again seasonally.
Why is Cloud Cost Optimization so controversial? Because most organizations perceive it as something to do after migrating and transforming (in other words, an operations function) …not as a something that is a fundamental and central part of the process of change. And in our experience, once the focus of Cloud Cost Optimization is removed from the Cloud Enablement Program, it tends to get lost in the noise of the next set of migrations and other enterprise priorities. In other words, it simply doesn't get done. After all, no one likes to diet…As Julia Child once famously said, “The only time to eat diet food is while you're waiting for the steak to cook.”
And to that end, we see great success with the following rules of Cloud Cost Optimization:
· Cloud Cost Optimization is more than a tool. It is an actionable and measured result that is a fundamental part of the transformation lifecycle.
· The CCoE should be responsible for the facilitation and execution of Cloud Cost Optimization as well as be held accountable for the results
· Active executive sponsorship is crucial to highlighting expense reductions and business units who are resistant to driving out waste.
· Cloud Cost Optimization is not a one-time function. It should be done routinely on a scheduled interval (weekly, monthly, quarterly)
· Don't miss macro savings by focusing only on micro items
· The CCoE should be viewed as an internal profit center and supported with investment in their ability to help continually drive out costs
In the end, if Cloud Cost Optimization isn’t made a high priority item, we see a strong chance of the cloud initiative failing in the long term. By leveraging and empowering the CCoE to drive Cloud Cost Optimization, you take most of the complexity and pain out of optimizing the cloud. As Brian Wansink, author of Mindless Eating: Why we eat more than we think once said, “The best diet is the one you don't know you're on.”