interLink: Offloading Kubernetes Pods to Anywhere
Shriira Press
How Kubernetes learns to run pods on supercomputers, HPC clusters, and any system it cannot reach on its own.
Welcome to interLink: Offloading Kubernetes Pods to Anywhere.
interLink is a CNCF Sandbox project that lets a Kubernetes cluster execute pods on remote resources it could never normally manage — Slurm and HTCondor batch systems at HPC centers, virtual machines running Docker or Podman, other Kubernetes clusters, even serverless backends. It does this by presenting a virtual node that looks ordinary to the scheduler but, behind the scenes, forwards each pod through a small REST API server to a pluggable backend that knows how to translate a container into a batch job or a remote run. This book starts with the problem of bridging cloud-native orchestration and heterogeneous computing, then builds up the Virtual Kubelet foundation interLink stands on, its three-piece architecture of virtual node, API server, and plugin, and the three deployment topologies that fit clusters, edge nodes, and tunneled multi-user setups. From there it covers the plugin REST contract that makes new backends easy to write, the practicalities of running real workloads through the Slurm plugin with its flavors and annotations, and finally where interLink fits in the wider ecosystem of scientific and AI computing.
This title is part of the ShriIra library and is free to read in full, right here — our small contribution to making world-class knowledge easy to reach.
A note on reading it: open the Contents menu at the top of the reader to jump between chapters, use the Aa menu to set a comfortable text size, theme (light, sepia, or night), and single- or two-page layout. Your place is saved automatically, so you can always pick up where you left off.
We hope it serves you well.
— Shriira Press