Cluster Challenge

The Cluster Challenge showcases the amazing power of clusters and the ability to harness open source software to solve interesting and important problems. Teams compete in real time on the exhibit floor, where they run a workload of real-world problems on clusters of their own design. The winning team will be chosen based on workload accomplished, benchmark performance and overall architectural design.

Building on the success of last year, the Cluster Challenge is featuring seven teams this year. The rules are simple: 26 amps (@110 volts), and up to six team members—no one can have a degree. The goal of the event is for teams to out compute one another by executing the HPC Challenge benchmarks faster and complete more of the application runs than the competition.

Teams assemble their clusters over the weekend prior to the Exhibit Hall opening. They are introduced and the race begins as the Exhibit Hall opens. For the next 44 straight hours, teams work toward the finish, when judges will arrive to view the results and interview the teams. The winner will be announced during the Awards Ceremony on Thursday afternoon.

Last year one team achieved 420 gigaflop/s on Linpack. Will we reach a teraflop this year, or more? Please plan to visit the Cluster Challenge area during Exhibit Hall hours Monday through Thursday and talk to the teams as they showcases how powerful and accessible clusters have become.

Team 1: Team Cluster Meister

The 2008 Olympics pits countries against one other in contests of speed and prowess. The 2008 Cluster Challenge features Technische Universität Dresden, Germany, and the Open Systems Lab of Indiana University, USA, combining forces to make a dash for the coveted Cluster Challenge Gold Medal. This year, our multinational team, fueled by bratwurst and soda, promises to keep you on the edge of your seat as they compute faster than an Indy car drives on the Autobahn.

Team 2: University of Alberta

Team University of Alberta, from Edmonton, Alberta, Canada, is led by students and faculty from the department of computing science. We were excited to win the Cluster Challenge in 2007, and we know the competition will be even stronger in 2008. But, we play to win. Therefore, once again, our vendor partner is SGI and our system is based on the latest revs of the Altix XE platform running Linux. The team's secret sauce is a mélange of excellent hardware, talented students, and coaches with strong opinions but with a hint of the ability to agree to disagree.

Team 3: Arizona State University

Arizona State University, partnered with Microsoft Corporation and the Arizona State University High Performance Computing Initiative, bring a diverse and dedicated team to the competition. We will stand out with the latest in cluster technology, including a brand new Intel-based blade system, DDR Infiniband, and the never-before-entered Windows HPC Server 2008 platform. We also plan to engage the audience with novel audio and haptic data interfaces. Watch out for the blazing blades of the Sun Devils!

Team 4: University of Colorado

From the beautiful Rocky Mountains comes Team CU. A team of six University of Colorado undergraduates in conjunction with industry leader Aspen Systems have designed a high performance Intel computing cluster which combines the best of commercial and open source technologies modified to meet the specifications of the Cluster Challenge. The University of Colorado's mascot, Ralphie the Buffalo, will be leading our team’s charge to win this year’s Cluster Challenge competition!

Team 5: National Tsing Hua University

Team NTHU, from National Tsing Hua University, comes from Taiwan. Our vendor partners are Hewlett-Packard Development Company (HP) and Intel Corporation. We have designed a system based on HP's BL2x220c double density blades with special low-power configuration running a customized Linux kernel. With the latest CPU provided by Intel, we have connected 80 3-Ghz cores with HP’s high bandwidth InfiniBand network in our system. Due to the super power of our cluster and our efforts, we believe that the final victory will come to us.

Team 6: Massachusetts Institute of Technology

Team GPU from MIT believes that Graphics Processing Units (GPUs) will be a key component in the design of future HPC clusters. We have designed a system that takes advantage of the special purpose processing capacity of the GPU and combines that with the traditional general processing capabilities of the CPU. We have made some unique heat removal modifications to lower our relatively high watts-to-flops ratio.

Team 7: Purdue University

If the 2008 Cluster Challenge was on St. Patrick’s Day, the “greener” entry from Purdue University, along with vendor partner SiCortex, would be right at home. The Purdue/SiCortex system is well suited for power-conscious HPC, potentially providing more flops-per-watt than conventional systems—nearly a teraflop of computation for less than 4 kilowatts of power. The system boasts 972 567-MHz, 64-bit MIPS processor cores and a unique high-speed interconnect fabric for parallel communication that can make it as capable as more power-hungry clusters.

Team members come from Purdue's College of Technology, Computer Sciences Department, and Electrical and Computer Engineering School as well as domain sciences. The team brings together undergraduate students being educated in all facets of high performance computing, from hardware design to parallel algorithms and programming to best practices in designing and operating HPC cluster systems. The 2008 Boilermaker Clustering Team looks forward to competing at SC08 in Austin and leaving the competition green with envy.