April 7, 2025

How to Balance Cost and Performance in EKS

Marie Jaksman
Growth Marketing Manager

Most Kubernetes teams play it safe — by overprovisioning. They assume more CPU and memory means better reliability. But in Amazon EKS, that usually just means higher bills.

The truth? You can tune for performance and cost — if you treat architecture like a balancing act. That’s the core message of the Frugal Architect webinar with AWS Senior Solutions Architect Nataliya Godunok.

“Performance and cost are not either/or — you can optimize for both over time.”

The Problem: Engineers provision "just in case."

When engineers don’t have clear data on how apps behave in production, they tend to provision too much “just in case.”

Default settings (like 512Mi memory or 1 vCPU) seem safe — but at scale, those small overprovisions add up fast.

You end up paying for resources your workloads barely use. And because it’s hard to trace spend back to specific apps or teams, no one feels ownership over cost.

The Solution: The Frugal Approach to EKS Optimization

1. Use Metrics for Performance Tuning

EKS optimization

“Granular observability is essential for making smart cost decisions.” - Nataliya Godunok, AWS Senior Solutions Architect

Collect real-time data on CPU and memory usage. Compare requests vs actual usage to identify overprovisioned containers.

Tools to use:

  • AWS CloudWatch Container Insights
  • OpenTelemetry
  • Prometheus + Grafana

2. Set Budgets Based on Business SLOs

Instead of guessing how much to allocate, tie performance goals to real outcomes.

  • Does the service need to scale during a product launch?
  • What’s the acceptable response time under peak load?

Use this to guide auto-scaling thresholds, not just usage spikes.

Frugal Architect
“Make cost a non-functional requirement.” - Law I: Putting it all together for continuous improvement

3. Create Small, Lean Container Images

“Only use the resources you need.” - Nataliya Godunok, AWS Senior Solutions Architect

Large container images slow down startup times and consume storage unnecessarily. By reducing image size:

  • Deployments are faster
  • Network egress is lower
  • Cold starts are shorter (especially in serverless environments)

Pro tip: Use multi-stage builds and strip unused dependencies.

EKS Optimization Common Pitfalls to Avoid

MistakeWhy It Hurts
Treating defaults as “safe” Leads to silent overprovisioning
Measuring infra onlyMisses app throughput and real-world performance
Ignoring long-term optimization“Set it and forget it” wastes money over time

EKS Optimization Metrics to Track

MetricSuccess Signal
Average container image size Should trend down over time
CPU/Memory usage vs requestCloser match = better efficiency
Cost per performance unit (e.g., requests/sec per $)Higher = better ROI

The Frugal Architect doesn’t chase perfection. Instead, it encourages regular tuning, like cleaning up old deployments or adjusting resource requests monthly.

“Architecting is a series of trade-offs.” - Nataliya Godunok, AWS Senior Solutions Architect

Furgal Architect


And you can’t make trade-offs if you don’t have data and ownership.

What's Next? 

  • Audit your container requests vs usage
  • Define SLOs that connect business outcomes to performance
  • Shrink your base images
  • Review metrics monthly, not just after incidents

Resources

PerfectScale Lettermark

Reduce your cloud bill and improve application performance today

Install in minutes and instantly receive actionable intelligence.
Subscribe to our newsletter
How to balance cost and performance in EKS? This blog breaks down insights from our Frugal Architect webinar with AWS's Nataliya Godunok.
This is some text inside of a div block.
This is some text inside of a div block.

About the author

This is some text inside of a div block.
more from this author
By clicking “Accept”, you agree to the storing of cookies on your device to enhance site navigation, analyze site usage, and assist in our marketing efforts. View our Privacy Policy for more information.