Grow/Shrink and Pre-Emption Policies Using Windows HPC Server 2008 Tutorial
< Go Back
This tutorial shows the power and ease-of-use of the grow/shrink and pre-emption job scheduling policies in Windows HPC Server 2008. These policies help the IT administrator manage job submission and execution, and optimize cluster resource usage.
With the grow/shrink policy, jobs can be dynamically sized over time to decrease queue wait time and job run time. This tutorial shows how, in Windows HPC Server 2008, a job can "grow" to use resources as they become available: If a node is added to the pool of available resources, the job can grow to take advantage of the new resources. A job can also "shrink" as it nears completion by releasing resources it no longer needs, making them available for other jobs. These policies make it easy for the administrator to manage the job queue and optimize the cluster resources.
With pre-emption, high priority jobs can take resources away from (or bump) already-running, lower-priority jobs. Learn how Windows HPC Server 2008 supports the pre-emption of jobs currently being executed in order to run higher-priority jobs. If a higher-priority job is placed in the queue, it executes before lower-priority jobs pending in the queue. See how the priority of a job can easily be changed while a job is running; graceful pre-emption lets tasks complete before interrupting the lower-priority job for the higher.
