Platform Computing Pursues Simpler Cluster Management

• By Dian Schaffhauser
• 05/23/11

High performance computing has always been a specialty area exploited by researchers and managed by IT administrators with arcane knowledge of how to work with clusters of computer processors. Platform Computing, which has been building and selling distributed computing systems since 1992, is looking to make the use of HPC more accessible with the release of a new version of its management software. Platform HPC 3 comes with a Web interface specifically redesigned to simplify monitoring and management of cluster environments. A cluster is a set of processors that can be linked to work together on a computing problem.

According to the company, the application provides features that are integrated to allow non-specialized users to deploy, run, and manage HPC clusters while still maintaining application performance and workload balance. The base product includes:

• The ability to provision and manage clusters;
• A workload scheduling engine to allow users to access and share computing resources more efficiently;
• A dashboard for workload and system monitoring, troubleshooting, and reporting;
• The ability for multiple operating system versions to be run concurrently on the same physical resources;
• The capability to schedule graphics-intensive applications to run on processors with graphics processing units (GPUs); and
• Support for multiple flavors of MPI--the message passing-interface--to expedite communication between nodes.

The latest release features updates to the interface, monitoring capabilities, availability, and cluster management itself. A new wizard walks the user through setup of provisioning. Cluster monitoring now encompasses GPUs, provides beefed up alerts and provisioning status updates, and integrates with third-party management software to watch over non-server devices. Failover for workload and cluster management, as well as reporting and monitoring tools, has been automated, as has automatic recovery when primary nodes return. Also, the new release includes improvements to cluster management, such as enhancements to kit-building tools for measurement and analysis of program performance and link aggregation.

"An increasing number of users are taking the management of their computing resources into their own hands, particularly in disciplines where projects have fast turnaround times and both productivity and cost cutting are key," said Tripp Purvis, vice president of business development at Platform Computing. "Platform HPC provides a complete management solution for the HPC community reducing the need for an in-house HPC expertise."

Platform HPC has been used in projects at Harvard Medical School, the University of Alabama, and Lancaster University in the United Kingdom.