Container vCPU Capacity Counting Methodology
Catalog row in Vendor License Evidence · Cited
- Kind
- Policy excerpt
- Excerpt
- The vCPU capacity of a pod is the sum of the CPU limits for all containers within that pod (Kubernetes resources.limits.cpu). Pods with vCPU capacity less than one core will be counted as fractions. Within a worker node, when the total vCPU capacity for all pods of an IBM Program exceeds the worker node capacity, the vCPU capacity is equal to the worker node capacity. vCPU capacity for an IBM Program will be aggregated at the cluster level and any fractional values rounded up to the nearest whole integer. Available vCPU capacity will be polled at frequent intervals (≤ 30 minutes) throughout each day; the maximum value from intraday calculations determines daily available vCPU capacity. Container HT/SMT: 2 vCPUs per 1 core as 1 VPC or 70 PVUs if Hyperthreading or SMT2; SMT4=4:1; SMT8=8:1; mixed SMT → lowest factor applied.
- Locator
- Container Licenses — vCPU Capacity Counting Methodology; HT/SMT; polling
- Applies to
- IBM Programs / Cloud Paks in containers under Container Licenses
- Aliases
- pod CPU limits; worker-node cap; cluster round-up; ≤30 min poll; container SMT factors
- Supports rules
- Related metrics
- Related programs
- Sources