Prerequisites and limitations
Prerequisites
Before implementing the strategies in this guide, ensure you have:
Technical requirements:
An active AWS account with Amazon EKS clusters running
Familiarity with Kubernetes concepts (Pods, Nodes, Deployments, DaemonSets)
Karpenter v0.33+ or Cluster Autoscaler installed on EKS clusters
Access to Amazon CloudWatch for metrics and monitoring
kubectl configured for EKS cluster access
Organizational requirements:
Executive sponsorship for cost optimization initiatives
Defined process for testing changes in non-production environments
Communication plan for workload migrations and disruptions
Cost allocation and chargeback model (recommended)
Tools and utilities:
Helm 3.x for installing Kubernetes applications
AWS CLI version 2.x
Monitoring tools (CloudWatch Container Insights, Prometheus, or similar)
Cost visibility tools (AWS Cost Explorer
, Kubecost, or similar)
Limitations
Be aware of the following limitations:
Technical limitations:
Karpenter requires Amazon EKS version 1.23 or later
Graviton migration requires multi-architecture container images
Some third-party software may not support ARM64 architecture
Spot instances may be interrupted with 2-minute notice
VPA and HPA cannot target the same metrics simultaneously
Operational considerations:
Node consolidation may cause temporary pod disruptions
Cluster upgrades require careful planning and testing
Some optimizations may require application code changes
Initial implementation requires time investment from engineering teams
Service quotas:
Amazon EC2 instance limits per region
Amazon EKS cluster limits (100 clusters per region by default)
Auto Scaling group limits
Spot instance capacity availability varies by region and instance type
