Rising Kubernetes Costs Highlight the Need for Better Monitoring and FinOps Optimization

The CNCF’s latest report suggests good news: Effective cost visibility tooling and FinOps practices do lead to cloud cost optimization. How to get there? That’s another story.

The latest Cloud Native and Kubernetes FinOps Microsurvey from the Cloud Native Computing Foundation (CNCF) found that cloud and Kubernetes adoption are breaking new records. Yet, Kubernetes-related cloud spending (and spend inefficiency) is rising just as quickly. While organizations have become more intent on establishing the cost insights and optimization necessary to rein in those bills, the CNCF’s report makes it clear that many have a way to go.

The report is worth reading in its entirety, but here are three key takeaways:

1. Kubernetes costs are rising, and overspending is increasingly common.
Forty-nine percent of survey respondents have seen cloud costs jump since they adopted Kubernetes, with 17% reporting “significant” increases to their bills since turning on the Kubernetes spigot.

However, Kubernetes commands vastly different chunks of cloud budgets across organizations. For half of the respondents, Kubernetes accounts for less than a quarter of their cloud spending; in contrast, 15% spend over half of their entire cloud budgets on Kubernetes alone.

Putting those costs into hard numbers: 21% of respondents spend less than $10,000/month on cloud-related costs, 26% spend under $50,000/month, and, at the high end, 22% of respondents are dropping more than $1 million each month on cloud bills.

Turning to scale, half of the organizations surveyed reported that they operate fewer than 50 nodes. Those operating at higher scales trail off from there, with 17% operating 251-500 nodes, and just 6% each in the categories of 251-500, 501-1,000, and over 1,000 nodes.

Comparing those numbers offers an immediate conclusion: The arcs of reported cloud spending and Kubernetes scaling don’t match. That strongly indicates widespread inefficiency.

Many who pay more aren’t getting more. When asked about the factors leading to this discrepancy, 70% of respondents cited over-provisioning as a top issue, easily leading the way as the top cause of overspending. Forty-five percent named a lack of individual or team-level responsibility for cloud costs. Two additional issues were each cited by 43% of respondents: Sprawl (such as resources not deactivated after use) and technical debt from existing workloads that aren’t re-architected to take advantage of cloud environment scalability.

2. Cost monitoring reflects the cloud marketplace, but remains vastly underutilized.
Organizations’ challenges in controlling cloud and Kubernetes costs are reflected in their use of cost-monitoring strategies. Thirty-eight percent of respondents confessed that they have no Kubernetes cost monitoring in place. Four in ten rely on their own estimates, and only 19% of organizations are equipped with accurate showback capabilities that enable their teams to understand the Kubernetes costs they’re responsible for. Even more concerning: Only 2% have an active chargeback program to actually enforce that responsibility.

Regarding specific Kubernetes cost monitoring and budgeting tools, responses largely aligned with the market shares of the hyperscale cloud providers. Many respondents, 55%, use AWS Cost Explorer, followed by GCP’s cost tools at 28% and Azure’s cost management at 23%. Kubecost was also the tool of choice for 23% of respondents, tying Microsoft and ahead of multi-cloud cost monitoring providers. OpenCost, the open-source CNCF Sandbox project for real-time Kubernetes cost monitoring that counts AWS, Google Cloud, Red Hat, Kubecost and many others among its founding contributors, came in at 11%.

Interestingly, the survey also asked respondents what would help them get overspending under control. Sixty-eight percent of respondents see advantages in making teams and individuals more aware of the need for spending discipline. Fifty-eight percent agreed that improved collaboration and communication on consumption and spending would rein in costs. Another 58% saw value in honing best practice blueprints that individuals and teams could consistently follow.

3. FinOps optimization is the clear goal.
FinOps — instilling shared cost responsibility across an organization — is a common objective for businesses needing more effective cloud and Kubernetes cost controls. Ninety-eight percent of respondents cited the importance of engineer, developer and product team involvement in cost controls and increased attentiveness to spending.

Organizations are optimistic: 75% expect to achieve full buy-in from these teams as active and effective cloud budget hawks.

When adopting the FinOps tools and practices that enable teams to take direct responsibility for their budgets, organizations are at all different stages of their journeys. Ten percent have yet to begin, 35% are currently at the stage of researching and evaluating FinOps budget analysis and forecasting tools, 20% have operationalized tooling and practices to the degree that they actively track, report and evaluate cloud consumption and spend, and 18% have reached the forefront of FinOps integration (and now continuously optimize their cloud utilization across teams and departments).

Controlling Cloud and Kubernetes Costs

Organizations that complete their cloud transformations — as well as those born in the cloud — realize that any migration or scale project triggers a new challenge of cost-optimizing those environments. The good news is that effective cost visibility tooling and FinOps practices do lead to cloud cost optimization, as the vanguard of survey respondents proves. Organizations with more work ahead before reaching that goal should take encouragement from those leaders’ example.