HFSS15: What a Scheduler Does

• Enables effective/efficient utilization of cluster’s resources consistent with organization’s goals

• Maintains queue(s) of jobs

• Maximizes throughput of the jobs by processing all jobs as fast as possible

• Typically, one job per cpu policy

• Allows choice of various scheduling policies (e.g. First Come First Serve, Priority Based, Preemption)

• Provides a suite of tools or utilities (graphical or command line) for end user to submit jobs, monitor jobs, abort jobs, suspend jobs, …

• Manages a compute cluster by running various interacting ‘services’ on head nodes and compute nodes

• Provides a programming interface to access ‘services’

Scheduler Managed Compute Cluster

Head node(s) typically maintains queues. Compute nodes are typically on a high speed network, to improve scalability of parallel jobs. Services running on nodes interact with each other to manage resources. End user tools communicate with services to submit/abort/suspend/etc. jobs.

Related Topic

High Performance Computing (HPC) Integration

Command Line Enhancements for Ansoft Desktop Products