Strategies for dramatically cutting public cloud costs [Q&A]
IT teams are increasingly moving to the public cloud for its supposed low cost and high agility. However, once implemented in their environments, people are finding public cloud to be expensive and fairly complex to manage.
We spoke to Andrew Hillier, CTO of cloud optimization analytics service provider Densify, who is all too familiar with the struggles that IT organizations face as their public cloud costs rise through the roof. Read on for his insights into strategies that organizations can implement when looking to reduce cost, risk and complexity in the public cloud.
BN: Many organizations today are finding issues with escalating public cloud bills. What's behind this? Is it possible to lower these costs without compromising application performance in cloud environments at the same time?
AH: Many IT professionals assume that public cloud is cheap, but that isn't always the case. In fact, we hear more often than not that organizations are over-running budgets by significant amounts -- a 50 percent over-run isn't uncommon from my experience. The good news is that there’s a huge opportunity to drive optimization in public cloud because a lot of the waste and excess comes from some pretty rudimentary approaches to cloud resource management. There are multiple strategies that can be employed, from the most basic of simply understanding what you're using and turning off instances you don't need, to properly right-sizing and modernizing instances to more advanced strategies -- all of which will impact your spend.
BN: Since you mentioned strategies, what can companies implement when looking to optimize their public cloud capacity at a lower cost than they're currently incurring?
AH: Optimizing public cloud resources is best approached in a tiered strategy. First and foremost, you should simply get visibility into what you are buying and what's running out there. Getting a clearer picture of what is contained in your bills will provide better insight into where your money is going so you can best determine how to get it under control.
To go beyond this is where most organizations -- and available tools -- fall short, as it does require detailed analysis of the workload patterns of the applications running in the cloud. The opportunities for optimization come from making better cloud instance selections from the vendor's catalog which includes right-sizing, moving between classes or looking at newer versions of an instance class, and identifying and removing deadwood.
We often see companies doing a pretty poor job of this, particularly when it comes to modernizing cloud instances to take advantage of the latest services and catalogs, which can often double your cost savings over right-sizing alone. The final tier of optimization can come from advanced hosting strategies such as stacking containers within a cloud instance to take advantage of resource over-commit. We have seen savings of up to 80 percent by doing this, although these savings tend to vary by type of workload.
BN: What advice would you give to organizations looking to right-size their environment based on their workloads?
AH: When preparing to right-size your environment, it's important to properly analyze the detailed intraday application workload patterns of each instance, and ideally do this for a full business cycle. This will give you real visibility into what resources an instance requires, when it peaks and for how long, and whether systems that appear to be idle are truly deadwood or not.
Many organizations rely on basic approaches, using simplistic models of peak and average workload utilization, to determine resource requirements, and end up with either resource shortfalls or huge waste. Analyzing the history of a workload to learn its pattern, and using this to predict future requirements, lets you precisely match workloads against all available instance types in order to find the lowest cost option that meets its needs.
Ideally this process is also done in concert with evaluating newer instance versions that are available, or even comparing options across instance classes, which gives you the opportunity to modernize your instances and usually save more money. To do this though, you need to be able to analyze across different hardware platforms with different specifications. This kind of analysis requires using benchmarks to normalize the workload data so you can accurately size and select the right instances. Ultimately, right-sizing comes down to having detailed workload analytics that will allow you to understand both your workloads and the capabilities and costs of the cloud services you want to buy.
BN: It seems as though cloud providers are constantly offering new services and changing pricing to add to the challenge. How can companies keep up with the latest services and catalogs without having to manually keep tabs on every single new offering?
AH: It's true, vendors like Amazon are notorious for changing their pricing and, along with that, they are always modernizing their offerings and releasing new services. It would pretty much be impossible to keep up with it without the help of software that can absorb the catalogs and costs and do the analysis for you.
Like cell phone or cable providers, these companies won't necessarily notify you when a better deal exists for your application workloads. It's usually up to the buyers to dig around and figure that out for themselves, but the offerings are so plentiful and the decision is so complex, humans can't do it alone.
There are a few tools out there that will keep the catalog up to date, but you also need that detailed workload pattern analysis to actually make any real use of it. Organizations also need to think about what kinds of skills they want to build in house. Do you want to pay a team of people just to keep up with pricing changes for AWS? Or is that something that is better outsourced to a team of experts that do this for a living and have the analytics to handle the changes and guide the decisions? Our customers breathe a sigh of relief when they learn that we offer a service that combines our analytics with expert advisors because it removes that burden from in-house teams.