CAP Sizing

Pretext

This wiki summarizes the results of baseline metrics collected for CAP deployments on various platforms. The activity can be tracked here. The following metrics were tracked:

• For Nodes: CPU(%), Memory(%), Memory(MiB), Disk Space(%)
• For Pods: CPU(%), Memory(MiB)

The setup was HA with credhub and autoscaler enabled.

Platforms

• EKS

  • Configuration:

    • Number of instances: 3
    • Instance Type: m5.large instances
    • vCPUs: 2
    • Memory: 8G
    • Volume: 60 GiB(gp2)

  • Metrics for

    • Nodes
    • Pods

• AKS

  • Configuration:

    • Number of instances: 3
    • Instance Type: Standard_DS4_v2
    • vCPUs: 8
    • Memory: 28G
    • Volume: 60 GiB(managed-premium)

  • Metrics for

    • Nodes
    • Pods

• GKE

  • Configuration:

    • Number of instances: 3
    • Instance Type: n1-standard-4
    • vCPUs: 4
    • Memory: 15G
    • Volume: 60 GiB(standard)

  • Metrics for

    • Nodes
    • Pods