Our story goes to show that you don’t have to be a decades-old business to benefit from modernization—and that there might be far more room for optimization in your cloud infrastructure than you realize. Learn how we identified inefficiencies in our DevOps practices and implemented a proactive Kubernetes cost monitoring strategy that saved us hundreds of thousands in annual cloud expenses.
See also: How to Do Multicloud Cost Optimization Right
Born in the Cloud, But Not Without Challenges
Our company was founded in 2017. We were ‘born in the cloud’ and reaped all of the advantages that come with that, right from day one. However, a few months ago we acknowledged that some inefficient DevOps practices were increasingly holding us back from the development pace we believed we could otherwise achieve. Habits such as hot patches, shifting code between servers, and needing to back-patch edits in the production environment to lower environments were a few of the processes impeding development progress. To get our DevOps in line, we adopted Kubernetes and used it to implement CI/CD, Infrastructure-as-Code (IaC), and well-maintained lower environments, among other DevOps best practices.
We had several congruent goals around our Kubernetes adoption. Our usage of Amazon EC2 instances was quickly increasing, making more efficient utilization a priority. Our engineering team was also particularly interested in measuring the improved cost efficiency achieved by the move to Kubernetes, and wanted the ability to precisely determine our cost of goods sold (COGS). We recognized that gaining complete visibility into the infrastructure costs of our products would enable new opportunities to manage our customer pricing and margins more strategically.
Why We Chose an External Vendor for Cost Monitoring
When exploring our prospects for implementing and managing such detailed cloud and Kubernetes cost monitoring, however, we quickly learned that tapping an external vendor solution would be far more affordable and feasible than hiring internal staff to build requisite capabilities. Our engineering team then proceeded to vet available technologies but found that, in many cases, the cost of a vendor solution was as high as our estimates for adding internal staff. Ultimately, we selected Kubecost as an open core option that we could self-host and that offered licensed support and the depth of features we were looking for.
Within one month we had our new monitoring platform implemented and reporting on Amazon EC2 costs from all of our accounts (and allowing us to visualize all that spending on a single pane of glass). This strategy gives us complete visibility into our infrastructure and costs, with all of our workloads tagged and attributed to the correct products or development projects. These real-time metrics now make it simple for our developers and DevOps to immediately recognize whenever workloads are poorly-sized or overprovisioned—and react accordingly. Whereas wasteful spending could go unnoticed until the cloud bill arrived and unattributed workloads could make it hard to assign responsibility for costs, we now have all the tools we need to track down and eliminate inefficiencies from the moment they begin.
The Benefits of Real Time Visibility: Achieving Significant Cost Savings
We leverage these new capabilities proactively and aggressively. This strategy has transformed our understanding of our cloud and Kubernetes spending, and led us to realize transformative savings for our business. Four months after putting our new cost monitoring platform in place, we recognized and clawed back $750,000 in annualized Kubernetes-related cloud costs.
Perhaps needless to say, adopting this approach to clamp down on spending and optimizing our cloud and Kubernetes efficiency has already surpassed our expectations. Not long ago, our FinOps team would regularly ask our engineers if anything could be done about the large AWS bills we had coming in. Now, we’re making sure every dollar is accountable and transparent. We’ve also achieved the ability to estimate our COGS with a high degree of accuracy, allowing us to optimize product margins and plan much better informed and more ambitious revenue targets.
See also: How Understanding Our Kubernetes Cloud Costs Enabled Us to Scale
A Call to Action: Prioritize Cost Visibility
We’re also just getting started, with goals to further optimize our cluster efficiency and confidence in our ability to scale our clusters (and cost monitoring) going forward. As we see it, that $750,000 in yearly savings we’ve already realized is only the beginning. For any company that lacks clear insights into their cloud and Kubernetes spending, I’d strongly recommend prioritizing change. You can point to our story as proof that you might just find hundreds of thousands of reasons for doing so.
Peter Bierfeldt is the CISO at Accrete AI, an enterprise AI software provider. He joined the company from Reltio, where he served as CISO and Head of Compliance.