Windows HPC Server 2008 R2 is addressing the increased computing demands of the market by providing fast and reliable management of clusters with Out-of-the-box support for thousands of processing cores. To optimize the productivity of these systems of scale R2 will include improved diagnostic, troubleshooting and reporting tools, new policies and better scheduling of interactive service workloads, and continued scale and reliability for service oriented HPC workloads.
Windows HPC Server 2008 R2 helps system administrators create an integrated, highly scalable, high performance computing (HPC) platform that can be deployed and managed with ease, and enables end users to efficiently access HPC resources from the familiarity of their desktop. Improvements in Visual Studio 2008 and in the upcoming Visual Studio 2010 release will assist developers to create parallel programs efficiently and integrate with the distributed resources offered by Windows HPC Server 2008 R2.
Job Scheduler
The job scheduler scales to a larger number of nodes and processing cores, while boosting overall cluster utilization through more sophisticated scheduling policies. Some examples present in the Beta include:
Scheduling at Scale and improved task throughput
Ability to monitor Job Progress with greater granularity
Ability to set up Pre and Post Tasks
Runtimes
Runtime support has now been expanded to include more robust programming models as well improved support for jobs based on the SOA programming model. Examples present in the Beta include:
Fire and Recollect Programming Model
Fire and Recollect Programming Model WCF Broker Re-architecture to permit job throttling etc.
Greater interoperability with Java based jobs
Management
With Windows HPC Server 2008 R2, you get a scalable platform both from a deployment and system management perspective as well as extensibility of the reporting functions to ensure more productive system management. Some examples included in the Beta are:
Deployment at large node counts
Efficient Management at large node counts
Customizable Heat Map for better system wide monitoring
Extensible reporting to tailor HPC infrastructure reports for specific requirements
