Add AKS Autometic vs EKS Auto mode pod startup latency test scenario

[anson627]

This pull request introduces a new performance evaluation scenario to compare pod startup latency between AKS Automatic and EKS Auto mode clusters. The changes include creating a dedicated test scenario, configuring pipelines, and adding Terraform inputs for Azure and AWS. Key aspects of the implementation include leveraging the cri-resource-consume topology, using clusterloader2 as the test engine, and enabling detailed metrics collection.

New Test Scenario: AKS Automatic vs. EKS Auto Mode Pod Startup Latency

• Scenario Documentation: Added a markdown file outlining the test scenario, implementation options, technical details, and expected outcomes. The recommended hybrid approach combines existing infrastructure with a dedicated scenario for automatic clusters.
• Terraform Inputs:
  • azure.tfvars: Configures AKS Automatic clusters with multi-zone support, a dedicated node pool, and specific settings for the test.
  • aws.tfvars: Configures EKS Auto mode clusters with a VPC, subnets, security groups, and managed node groups for the test.
  • Test inputs (azure.json, aws.json): Provide region-specific configurations for local testing. [1] [2]

Pipeline Configuration Updates

• Pipeline YAML: Created a new pipeline configuration to execute the test scenario on Azure and AWS. Includes matrix configurations for pod densities, Kubernetes versions, and memory-focused workloads. Enabled kubelet scraping for detailed metrics.