Slurm and PBS Integration
Fuzzball integration with the Slurm and PBS batch scheduling systems enables you to leverage existing HPC clusters as compute resources for running Fuzzball workflows. This integration allows Fuzzball to submit jobs to Slurm/PBS, monitor their execution, and manage resources through these familiar workload manager interfaces.
The integration with Slurm and PBS helps you to do the following:
- Leverage Existing Infrastructure: Use your existing cluster without modification.
- Share Resources: Share cluster resources between traditional Slurm/PBS jobs and Fuzzball workflows.
- Use Familiar Administration: Continue using existing Slurm/PBS scheduling policies and tools.
- Maximize Efficiency: Optimize utilization of on-premises HPC infrastructure.
The Slurm and PBS integration is configured through two main mechanisms:
The Orchestrate configuration defines how Fuzzball connects to your Slurm or PBS cluster. This is
typically configured through the FuzzballOrchestrate Custom Resource Definition (CRD) in
Kubernetes deployments.
See Orchestrator Configuration for details.
Provisioner definitions specify the compute resources available through Slurm or PBS and how they should be allocated. These definitions are created through the central configuration system.
See Provisioner Configuration for details.