Problem description
The Edge-Application-Management's current RequiredResources' numCPU field is an integer, which restricts vcpu resource allocation to whole vcpus. Notably Kubernetes allows for fractional vcpu allocation down to the millivcpu, to allow for very lightweight containers/microservices that do not use a lot of compute.
Expected behavior
Being able to specify required vcpus in milli-vcpu increments.
Vcpu:
type: string
pattern: ^\d+((\.\d{1,3})|(m))?$
description: Number of vcpus in whole, decimal up to millivcpu, or millivcpu format.
example:
whole:
value: 2
decimal:
value: 0.500
millivcpu:
value: 500m
For VM-based deployments, fractional vcpus may not be allowed, but that should be implementation dependent. For example a VM deployed via Kata-containers (kube-virt) on a Kubernetes cluster may allow for fractional vcpus.
Problem description The Edge-Application-Management's current RequiredResources' numCPU field is an integer, which restricts vcpu resource allocation to whole vcpus. Notably Kubernetes allows for fractional vcpu allocation down to the millivcpu, to allow for very lightweight containers/microservices that do not use a lot of compute.
Expected behavior Being able to specify required vcpus in milli-vcpu increments.
Alternative solution None.
Additional context The GSMA Federation EWBI already allows for this:
For VM-based deployments, fractional vcpus may not be allowed, but that should be implementation dependent. For example a VM deployed via Kata-containers (kube-virt) on a Kubernetes cluster may allow for fractional vcpus.