Supercluster Discussion From IAC Meeting Wednesday, April 19, 2006 Denny George (Decisions) Research Challenges Inherently trustworthy systems, data, information and knowledge Transparency, mobility and global access/delivery Better hammers (algorithms, models, methods, ...) Research Themes Trusted Systems Trusted Systems: Impacting our abilities to avoid natural and manmade disasters Trusted Systems: Connecting and Sustaining Our World in the 21st Century Joan Mitchell (IBM) Bradford Brooks sent his regrets. He expected to mention that a challenge for the future is how to reduce the power requirements for a computer center. Tackling this problem will require interdisciplinary efforts at all levels. I noted to the group that I had actually seen a computer built with vacuum tubes. Todays requirements for power are much reduced from that, but the hunger for more servers and more processing is growing faster than the power reductions. With the energy crisis, the conservation of resources (e.g. power) becomes even more significant. The second grand challenge is how to move people up the value chain. Both Brad and I agree that the future of our country depends upon working professionals being able to get advanced degrees and retrain in new disciplines. This takes cooperation from Universities and Industry. The retraining is likely to be much less expensive than training people from scratch. Keith Mathias (Northrup Grumman) We are very excited about the SuperCluster proposal because it appears to be attacking the right problems, will produce some very important technology and will result in some top-notch recruits. Programs like this will benefit the industry, and society in total, by technology insertions via the students that have worked on these projects. There is no better way to insert technology than to send a capable person to assist in the insertion. The proposal addresses many of the hard problems that we are faced with in industry today. Research Challenges: Information management: Transformation of information into knowledge. This is regarded as a very difficult problem. How are the relevant features identified? Large volumes of data, organization of data, responsive queries, and low cost. Hardware and Algorithms: Building faster hardware is important but more efficient algorithms are likely to result in bigger gains. Improving search performance in HW or SW needs to improve by orders of magnitude not just by constant factors Network Modeling and Security Identity and access validation Performance (bandwidth and speed) Information warfare. These problems affect a wide range of industries from homeland security to financial information and business intellectual property.