EKS and Container Cost Optimization

Reduce cluster costs by right-sizing nodes and workloads. FCAI detects inefficient pods and suggests safe limits.

Cluster and node strategy

• Use managed node groups or Karpenter with Spot where appropriate.
• Prefer Graviton (c7g, m7g, r7g) for 20–40% better price/perf.
• Separate latency-critical and batch workloads into distinct node groups.

Pod requests/limits

• Set CPU/memory requests to P50 usage; limits to P95–P99 to prevent eviction.
• Enable VPA for batch/cron, use HPA for services.
• FCAI flags over-requested pods and proposes new values.

Scheduling and bin packing

• Use topology spread, taints/tolerations, and pod priority for efficiency.
• Consolidate small nodes into fewer larger nodes when possible.

Autoscaling

• Calibrate HPA targets (60–70% CPU, or custom SLI) for better packing.
• Use Cluster Autoscaler or Karpenter with consolidation enabled.

Observability

• Collect pod and node utilization; watch throttling and OOMKills.
• FCAI dashboards show wasted CPU/memory and savings estimates.