APPLICATIONS
Indiana University, Technische Universitat Dresden form challenge team
The "ClusterMeisters," a team of undergraduates from Indiana University and Technische Universität Dresden, will compete in the 2008 Cluster Challenge at SC08, the premier international conference for high performance computing, networking, storage and analysis. SC, formerly known as Supercomputing, celebrates its 20th anniversary when it opens November 17, 2008 in Austin, Texas. The goal of the Cluster Challenge competition is to showcase the ability of clusters and open-source software to solve interesting and important problems, and to expose undergraduates to state-of-the-art systems and the best and brightest of their peers. The winning team will be chosen based on benchmark performance, workload accomplished, and overall knowledge and presentation. Teams will assemble their computer clusters after arriving in Austin. The computational hardware must be commercially available, and in order to level the playing field, the rules include tight physical constraints on the systems: the cluster computing hardware (processors, switch, storage, etc.) must fit into a single rack powered by two 120-volt, 13-amp circuits. This also encourages "green computing" by rewarding computations performed per watt of energy consumed. Teams are encouraged to join with one or more vendors, who may support team activities with equipment, training, and financial support. The IU-TUD team is working with IBM and Myricom. IBM is providing a new System X iDataPlex server, designed to use up to 40% less power than similarly configured standard servers. The nodes will be equipped with the latest energy efficient Intel Xeon quad core processors and energy-efficient memory. This provides the team with a cluster that has already shown a performance of more than 1 TFLOP/s (1 trillion mathematical calculations per second). The networking equipment provided by Myricom uses significantly less power than other available high-performance networking technologies. On Monday, November 17, all teams will have 6 hours to run the HPCC benchmarks and submit their results. On Monday evening the "race" starts and students will run five widely used scientific applications on their clusters, attempting to process as much data as possible without exceeding the power constraint. The race ends Wednesday at 4 PM. The judges will then meet with each team, inspect results, discuss the cluster solutions and team strategy with the students. The winning team will be announced at Thursday's SC08 award ceremony. The team will be advised by Professor Andrew Lumsdaine of IU's Open Systems Laboratory, part of Pervasive Technology Labs at IU, and Professor Wolfgang E. Nagel of TUD's Center for Information Services and High Performance Computing (ZIH). The faculty advisors are educating their students about power consumption, multi-core multi-node systems, and performance optimization. The team is coordinating its work via intercontinental videoconferencing, until they meet in Austin. Professor Lumsdaine remarks, "We believe that in today's computer industry, global teams of experts are necessary to achieve highest performance. The Cluster Challenge will allow our students to demonstrate the acquired skills in a multi-national team that combines the best of both institutions. It will be an amazing learning experience in international cooperation and high performance computing for them."