Skip to main content

Infrastructure Performance Benchmark

📅 Written: 2026-02-13 | Last Modified: 2026-02-13 | ⏱️ Reading Time: ~1 min

We measure and analyze key performance metrics of EKS cluster infrastructure.

Network Performance​

Cilium ENI vs VPC CNI Comparison​

A detailed quantitative comparison of VPC CNI and Cilium CNI across multiple modes (kube-proxy, kube-proxy-less, ENI, with tuning applied) is covered in a separate dedicated document.

For detailed benchmark results, refer to CNI Performance Comparison Benchmark.

Comparison Scenarios (5):

  • A: VPC CNI baseline (kube-proxy + iptables)
  • B: Cilium + kube-proxy (Overlay)
  • C: Cilium kube-proxy-less (eBPF replacement)
  • D: Cilium ENI mode (Native Routing)
  • E: Cilium ENI + full tuning (DSR, XDP, Socket LB, etc.)

Gateway API Performance​

Metrics

  • Request processing latency (P50, P95, P99)
  • Requests per second (RPS) throughput
  • TLS handshake overhead

DNS Performance​

CoreDNS Before/After Optimization​

Metrics

  • DNS resolution latency
  • Cache hit rate
  • Queries per second throughput
  • NodeLocal DNSCache effect

Autoscaling Response Time​

Karpenter vs Cluster Autoscaler​

Metrics

  • Scale-out response time
  • Node provisioning time
  • Pod scheduling delay

Cost Efficiency​

Cost vs Performance by Instance Type​

Metrics

  • Throughput per vCPU
  • Cost per memory
  • Spot vs On-Demand stability comparison