Under the STAR programme, the two organisations expect to co-develop and potentially commercialize a next-generation HPC software stack. TACC researchers will work with members of Univa UD's product organisation to develop advanced, open source HPC systems and operations management software capable of supporting highly scalable infrastructures.

As an industrial affiliate of TACC, Univa UD will also receive a range of benefits from workshops and publications to knowledge transfer from TACC experts in advanced computing technologies.

"Working with TACC, we hope to develop an open source, commercially supported reference cluster management architecture for next-generation HPC clusters", stated Gary Tyreman, vice president and general manager of HPC at Univa UD. "The project will be the first of its kind addressing a complete integrated solution that combines the essential components for scalable, professional cluster administration."

Key areas of development and integration will include tools necessary to install, configure and manage an efficient HPC cluster, such as:

• Operations Management
• Batch Management
• Software Control Strategy
• Host Monitoring and Logging
• Hardware Performance
• Provisioning

TACC is home to "Ranger", one of the most powerful supercomputing systems in the world for open science research. Ranger is also the largest supercomputer on the TeraGrid, a National Science Foundation-sponsored network of academic HPC centrrs that provides researchers access to large-scale computing power and resources.

"We are excited to collaborate with Univa UD in this effort to enhance the very large-scale cluster management techniques that are required to ensure consistent performance and reliability across many thousands of processor cores", stated Karl Schulz, associate director of high performance computing at TACC.

"Increasingly, HPC system administrators are looking to take advantage of the network effect of commercially integrated and supported run-book and management methodologies that help reduce the cost of operating a cluster", stated Gary Tyreman. "This partnership will deliver innovative operations management software that simplifies key steps in the HPC lifecycle by automating costly manual operational processes."

Technology developed from this collaboration may be incorporated into UniCluster, a fully integrated, community-enabled open source cluster software stack with unrestricted support for all components. Available by download, this product simplifies the deployment, configuration and management of clusters from installation and job scheduling to monitoring and remote access.

UniCluster delivers on Univa UD's HPC mission of bringing systems management best practices to HPC while providing better value for a better price.

The UniCluster 4.0 Test Drive is now available for download at www.grid.org/downloads . This open source project will be maintained on Grid.org, an on-line community for open source cluster and grid software users, administrators and developers. The mission of the site is to provide a single location where open source cluster and Grid information can be aggregated so people with a similar range of interests can easily exchange information, experiences and ideas. In addition the project is open to any organisation that wishes to contribute to, or benefit from, the project's output.

Preview versions of the technology will be made available for download on www.grid.org , with new updates added monthly.