Peerd Capability for Preemptive Pulling of OCI Content

[johnsonshi]

From docs, customers that want to pull large OCI format images (think large LLM images with gigabyte weights) to a multi-thousand node K8S cluster have no intuitive way to "warm" up workload scale up. A K8S admin interested in this scenario can of course create 1 pod, pull the image to that pod (which causes the workload to begin executing on that pod), before scaling to more pods to leverage P2P.

However, a customer may be interested in pre-emptively caching OCI content on the cluster (1) without image pulls and (2) without the pulled image running on any pod because they are interested in all images running at the same time. This isn't easily done and requires a lot of manual K8S scripts to achieve.

If Peerd had an intuitive CLI command or a K8S YAML experience for "pre-warming", it would facilitate community adoption.