High-Performance Computing Workload Manager

A High-Performance Computing Workload Manager is a workload manager that can support high-performance computing workload management tasks.

• AKA: HPC Workload Manager, HPC Job Scheduler, HPC Resource Manager, Supercomputing Workload Manager.
• Context:
  • It can typically schedule HPC Parallel Jobs through HPC job schedulers and HPC resource allocators.
  • It can typically manage HPC Compute Resources using HPC node managers and HPC resource monitors.
  • It can typically implement HPC Job Priority via HPC fair-share algorithms and HPC priority queues.
  • It can typically support HPC Job Arrays through HPC array job schedulers and HPC task farmers.
  • It can typically enforce HPC Resource Quotas using HPC accounting systems and HPC billing mechanisms.
  • It can often enable HPC Checkpoint-Restart via HPC checkpoint librarys and HPC restart protocols.
  • It can often facilitate HPC Topology-Aware Scheduling through HPC network topology maps and HPC locality optimizers.
  • It can often provide HPC Power Management using HPC power cappings and HPC frequency scalings.
  • It can range from being a Cluster-Scale High-Performance Computing Workload Manager to being a Exascale High-Performance Computing Workload Manager, depending on its HPC system scale.
  • It can range from being a Batch-Only High-Performance Computing Workload Manager to being a Interactive High-Performance Computing Workload Manager, depending on its HPC job type support.
  • It can range from being a CPU-Only High-Performance Computing Workload Manager to being a Heterogeneous High-Performance Computing Workload Manager, depending on its HPC accelerator support.
  • It can range from being a Single-Site High-Performance Computing Workload Manager to being a Grid-Enabled High-Performance Computing Workload Manager, depending on its HPC federation capability.
  • ...
• Examples:
  • Open-Source HPC Workload Managers, such as:
    • Slurm HPC Workload Manager for scalable HPC workload management.
    • Open Grid Scheduler for grid-based HPC workload management.
    • TORQUE Resource Manager for traditional HPC workload management.
  • Commercial HPC Workload Managers, such as:
    • PBS Professional for enterprise HPC workload management.
    • IBM Spectrum LSF for IBM HPC workload management.
    • Altair Grid Engine for commercial HPC workload management.
  • Cloud HPC Workload Managers, such as:
    • AWS Batch for cloud-native HPC workload management.
    • Azure CycleCloud for Microsoft Azure HPC workload management.
  • ...
• Counter-Examples:
  • Container Orchestration System, which manages containerized applications rather than HPC batch jobs.
  • Workflow Management System, which coordinates business processes rather than scientific computations.
  • Desktop Job Scheduler, which handles single-machine tasks rather than distributed HPC jobs.
• See: Workload Manager, HPC Resource Scheduler, Cluster Management System, Scientific Computing Platform, Distributed Resource Control System, Batch Processing System, Parallel Computing Infrastructure.

---