Advanced Job Scheduling with Windows HPC Server 2008
< Go Back
This presentation provides an in-depth look at the new Job Scheduler in Windows HPC Server 2008. It begins with a look at Abe, the Windows HPC Server 2008-based cluster at NCSA that ranked at number 23 on the June 2008 TOP500 list. The discussion then goes on to examine the motivation for parallel high-performance computing (HPC) models, and the focus on interactive applications.
Next, the presentation covers job scheduler terminology and concepts, and then considers the benefits of a scenario that combines HPC and Windows Communication Foundation (WCF) services. These concepts are illustrated with a demonstration of parallel ray tracing using an HPC + WCF services model. Three key facets of job scheduling are then described in detail: admission, allocation, and activation. See an example of submitting and tracking a job using the C# API, as well as detailed demonstrations of Job Templates and the different scheduling policies involved in job scheduling. The presentation concludes with a brief discussion of security issues.
