Where are you using WebAssembly?
Wasm promises to let developers build once and run anywhere. Are you using it yet?
At work, for production apps
At work, but not for production apps
I don’t use WebAssembly but expect to when the technology matures
I have no plans to use WebAssembly
No plans and I get mad whenever I see the buzzword
FinOps / Kubernetes

Key Considerations for Containing Kubernetes Costs

By implementing the following techniques, you don’t have to sacrifice your budget for performance.
Feb 13th, 2024 6:32am by
Featued image for: Key Considerations for Containing Kubernetes Costs
Image from chenspec on Pixabay.

Kubernetes has become the essential backbone to millions of applications, relied upon by two-thirds of Fortune 100 companies and hundreds of thousands of applications. Gartner predicts that by 2027, more than 90% of global organizations will be running containerized applications in production. However, as the reliance on Kubernetes increases, so do costs. According to the CNCF, 28% of organizations report Kubernetes is taking up to half their budget.

But by implementing the following techniques, you don’t have to sacrifice your budget for performance. As the saying goes: You can have your cake and eat it too.

Rein in Your Resources

Resources, like memory and storage, are often shared between teams using the same cluster. But without visibility into everyone’s environments, you’re at risk of over-allocating your resources and overspending on your cloud bills.

But what if you could set controls to avoid overconsumption? Enter resource quotas. Resource quotas help prevent this kind of overallocation by setting precise memory and CPU limits per team or namespace within a cluster. Setting up a tailored guardrail like this helps optimize resource utilization and keeps it fair by ensuring no single team or application can consume more than the resources they’ve been assigned, saving you from busting your budget.

While resource quotas are great for managing the overall resource consumption per namespace, that alone won’t prevent overconsumption. Incorporating resource requests ensures that containers are guaranteed to have the necessary resources to run, while resource limits help to define the maximum amount of CPU and memory that containers within that namespace can use.

You can think of quotas as setting your total monthly budget, limits determine how much you can spend on a single purchase, and resource requests are the monthly bills (like groceries) that you always need to accommodate for. Together, these tools help ensure you have enough resources without unnecessary overconsumption, ultimately resulting in better cost efficiency within your Kubernetes clusters.

Autoscale Where You Can

Resource quotas, limits and requests are great first steps in containing Kubernetes costs, but they fall short of addressing the dynamic nature of resource demands. That’s why you also need to incorporate horizontal pod autoscaling (HPA) and vertical pod autoscaling (VPA) into your strategy.

HPA automatically adjusts the number of pod replicas based on application-specific metrics. This ensures optimal resource allocation and responsiveness to varying workloads, much like how you’d set a thermostat in your home so your heating system knows when to turn on and off in response to outdoor temperature fluctuations.

Conversely, VPA dynamically adjusts CPU and memory resources for individual pods based on their actual usage. Think of it as a tailor adjusting clothing to fit perfectly for each wearer. This responsive allocation guarantees resource efficiency without overprovisioning.

It’s not recommended to use HPA and VPA together, but if you’re looking to make autoscaling even more effective, you can incorporate predictive analytics. Just as retail companies hire more employees during busy seasons, predictive autoscaling uses historical data to forecast resource demands. Some of the more popular methods to predict CPU demand are Holt-Winters exponential smoothing and long short-term memory neural networks.

Autoscaling is a necessary piece of the cost optimization puzzle. By taking the burden off engineers to adjust your resources to meet demand, you’re saving time, money and resources.

Use Machine Learning

Anomaly detection is an important aspect of cost optimization. When you combine alerting with anomaly detection, you can quickly identify outliers in resource usage and resolve issues before they result in unexpected costs.

For example, if you see a pod has higher error rates compared to other pods in the same service, you’re able to investigate the root cause and take action to address the issue — whether that’s adjusting resource allocation or moving pods to different nodes in the cluster.

Optimize Storage Costs Dynamically

Dynamic storage optimization is another approach that reduces storage costs while maintaining data accessibility. By provisioning storage intelligently and implementing data life cycle management policies, dynamic storage optimization can increase storage volumes without an administrator getting involved. This automatic provisioning of storage as and when required can reduce storage costs and improve resource utilization.

Lay the Foundation for a FinOps-Focused Future

FinOps, as defined by the FinOps Foundation, is “an operational framework and cultural practice which maximizes the business value of cloud, enables timely data-driven decision-making, and creates financial accountability through collaboration between engineering, finance and business teams.” By involving more teams and incorporating structural changes into the way your company operates, you can elevate cost optimization into an organization-wide goal and better control Kubernetes spend while maintaining business continuity.

In practice, this means having regular reviews with engineering, finance and business teams to facilitate cross-functional conversations about where to invest and when. This also fosters a culture of knowledge sharing so that cost-efficient resource allocation practices can be mirrored in various parts of the business.

The more you can do to make your stakeholders aware of the financial impacts of their engineering decisions, the more likely it is that they’ll factor it into their future choices.

Monitor Your Progress

While the strategies outlined above should help keep Kubernetes costs in check, how can you be sure? And how will you know when you need to revisit and revise some of those strategies based on the ever-changing needs of your business?

Monitoring tools like OpenCost provide visibility into current and historical Kubernetes spend and resource allocation to give you granular insights into where, when and how you’re spending your money.

As Kubernetes continues to grow and evolve, so will your cost optimization journey. There will always be new opportunities to increase efficiencies and cut costs, but the sooner you start, the more you’ll save.

To learn more about Kubernetes and the cloud native ecosystem, join us at KubeCon + CloudNativeCon Europe in Paris from Mar. 19-22, 2024.

Group Created with Sketch.
THE NEW STACK UPDATE A newsletter digest of the week’s most important stories & analyses.