skip to: onlinetools | mainnavigation | content | footer


Jim Ang
Sandia National Laboratories
Albuquerque, NM
Phone: (505) 845-7018
Fax: (505) 284-2518

Jeffrey A. Nichols
Oak Ridge National Laboratory
Oak Ridge, TN
Phone: (865) 574-6224
Fax: (865) 574-6076

IAA - Goals

Key metrics for success of the Institute include furthering DOE's missions in science, national competitiveness and national security. Successful impact on U. S. Computer Industry will be measured through changes to product roadmaps, and integration or adoption of IAA-sponsored technologies into next generation supercomputer systems. Impact on the DOE programs will be measured through the shift in priorities for ASC and ASCR programs in Algorithms, Applications and enabling technology development that creates synergy for the investments that the IAA is making in advanced architectures and systems. Impact on other agencies will be measured through the ability to attract and create synergy with coordination of investments in next generation computer architectures. Opportunities exist for coordination with DARPA-HPCS, NSF, and DoD.

Impacting Vendor Roadmaps. Ultimately, the IAA will succeed only by helping industry to incorporate technologies that are relevant to DOE applications into their roadmaps. Many vendors realize that HPC represents the vanguard of high-end applications, and that national laboratories have the science and engineering capabilities to cultivate the needs of those applications. But without the coordination to link industry efforts in HPC architectures with national laboratory efforts in HPC applications and algorithms a gap will remain. Consequently, carefully targeted NRE investments can significantly impact future HPC systems. The establishment of the IAA provides a vehicle for continuous interaction with HPC vendors.

Fostering Industry Collaboration. Vendors typically operate on 2-3 year design cycles, while academics are generally not reuiqred to implement their proposed architectures to demonstrate success. The IAA has the unique ability to operate on longer time-scales while simultaneously producing real implementations that demonstrate an engineering concept. Consequently, the IAA will foster collaborations among industry, academia, and the national labs to address the shortfalls in HPC architecture research.

Developing Prototype Systems. Fundamentally, the only way to lower the risk of adopting novel architectures and technologies is to demonstrate through paper studies, system simulation, and hardware prototypes the performance benefit of these technologies. A final step in the IAA's multidisciplinary approach to computer architecture research will be deploying small prototype systems. These systems will facilitate application, algorithm and system software development, prove the technology to industry, and lower the risk of DOE adoption of advanced architectures.

Creating and Deploying Software and Tools to the HPC community. Frequently, members of the HPC community need precise details for application workloads, architectural features, system software, performance and simulation models, and usage scenarios in order to assess design tradeoffs. Unfortunately, there are few widely accepted standards or tools for distributing this type of information. Therefore, to facilitate collaboration and interaction with HPC industry and academia, the IAA will create and deploy application requirements, benchmarks, software, performance prediction tools, and simulators.

Web site contact: