https://deploy-preview-846--kubernetes-contributor.netlify.app/community/community-groups/wg/batch/Recent content in WG Batch on Kubernetes ContributorsHugoenhttps://deploy-preview-846--kubernetes-contributor.netlify.app/community/community-groups/wg/batch/charter/Mon, 01 Jan 0001 00:00:00 +0000https://deploy-preview-846--kubernetes-contributor.netlify.app/community/community-groups/wg/batch/charter/<h1 id="wg-batch-charter">WG Batch Charter</h1> <p>This charter adheres to the conventions described in the <a href="https://deploy-preview-846--kubernetes-contributor.netlify.app/committee-steering/governance/README.md" >Kubernetes Charter README</a> and uses the Roles and Organization Management outlined in <a href="https://deploy-preview-846--kubernetes-contributor.netlify.app/committee-steering/governance/wg-governance.md" >wg-governance</a> .</p> <h2 id="scope">Scope</h2> <p>Discuss and enhance the support for Batch (eg. HPC, AI/ML, data analytics, CI) workloads in core Kubernetes. We want to unify the way users deploy batch workloads to improve portability and to simplify supportability for Kubernetes providers.</p> <h3 id="in-scope">In scope</h3> <ul> <li>To reduce fragmentation in the k8s batch ecosystem: congregate leads and users from different external and internal projects and user groups (CNCF TAGs, k8s sub-projects focused on batch-related features such as topology-aware scheduling) in the batch ecosystem to gather requirements, validate designs and encourage reutilization of core kubernetes APIs.</li> <li>The following recommendations for enhancements: <ul> <li>Additions to the batch API group, currently including Job and CronJob resources that benefit batch use cases such as HPC, AI/ML, data analytics and CI.</li> <li>Primitives for job-level queueing, not limited to the k8s Job resource. Long-term, this could include multi-cluster support.</li> <li>Primitives to control and maximize utilization of resources in fixed-size clusters (on-prem) and elastic clusters (cloud).</li> <li>Runtime and scheduling support for specialized hardware (GPUs, NUMA, RDMA, etc.)</li> </ul> </li> </ul> <h3 id="out-of-scope">Out of scope</h3> <ul> <li>Addition of new API kinds that serve a specialized type of workload. The focus should be on general APIs that specialized controllers can build on top of.</li> <li>Uses of the batch APIs as support for serving workloads (eg. backups, upgrades, migrations). These can be served by existing SIGs.</li> <li>Proposals that duplicate the functionality of core kubernetes components (job-controller, kube-scheduler, cluster-autoscaler).</li> <li>Job workflows or pipelines. Mature third party frameworks serve these use cases with the current kubernetes primitives. But additional primitives to support these frameworks could be in scope.</li> </ul> <h2 id="stakeholders">Stakeholders</h2> <p>Stakeholders in this working group span multiple SIGs that own parts of the code in core kubernetes components and addons.</p>