The Center for Computation & Technology, or CCT, at LSU has announced the hire of supercomputing expert Thomas Sterling. Sterling is leaving his position at the NASA Jet Propulsion Laboratory to continue his computing research in Louisiana. The computer scientist will arrive at the LSU campus on Aug. 22, accepting a professorship in the LSU computer science department.
Sterling says that LSU has the kind of environment that will allow him to contribute and collaborate. “Working at the CCT will give me a chance to continue to make an impact in high performance computing, and that is what is most important to me,” said Sterling. “It’s a place where creativity in computing is taking place.”
Sterling received his Ph.D. from MIT as a Hertz Fellow in 1984. He has worked in support of the President’s Information Technology Advisory Committee, and was a member of the DOD Integrated High End Computing Initiative. Sterling also participated in the High End Computing Revitalization Task Force. He has published many papers on computer-system architecture and holds six patents. He has authored and co-authored several books in the field of high-performance computing, including the influential books “Enabling Technologies for Petaflops Computing” and “How to Build a Beowulf.”
Sterling is best known for his work as leader of the Beowulf project, which he started in late 1993 at NASA. This world-famous project led to an efficient way of building high-end computers by networking off-the-shelf PCs. These computers, labeled Beowulf-class clusters, have demonstrated exceptional performance to cost advantages when compared to custom high-end computers. These clusters dominate the industry’s popular list of the world’s Top 500 Supercomputers at www.top500.org, and LSU’s SuperMike cluster was made possible because of this project.
Although the Beowulf model and the class of clusters it helped to create is the dominant form of high-end computer in use today, Sterling believes that it may no longer be the most effective way to build future generation supercomputers. He explains that “the great irony of his life” lies in the fact that Beowulf became so popular that it is now hard for researchers in his field to break away from that model and pursue innovative and superior approaches.
“We are trapped by the success of Beowulf and using the same paradigm as 20 years ago. You do have to make it cost effective, but the urban legends of computing are holding us back. Processors don’t have to be complicated,” he said, “a more simple architecture can be built that works better and more efficiently in a parallel computer.”
Sterling is now researching a type of “high density” machine called the “MIND” architecture, which stands for Memory, Intelligence and Network Device. He believes this design is a more innovative and effective way to build the next generation of high-performance computer.
One class of high-density system being developed by industry involves multicore chips. These computer chips each have multiple processors. IBM’s Blue Gene/L is a multicore architecture and is the fastest general-purpose computer in the world. The Sony Playstation 3, due out some time in the next year, also uses multicore high-density processor chips. This gaming device has nine processors per chip.
Sterling’s MIND architecture also uses a multicore chip, but this new design will achieve a higher degree of efficiency. “It solves problems of latency, scheduling and synchronization for parallel computing resulting in lower cost, lower power, and ease of programming not found on today’s high-performance systems,” said Sterling.
Dan Reed, Director of the Renaissance Computing Institute and recent chair of the computational science subcommittee of the President’s Information Technology Advisory Committee, has worked with Sterling in a variety of roles, including as part of the High End Computing Revitalization Task Force. Reed feels that an internationally known scientist like Sterling will move the CCT forward. “He brings a new dimension to CCT,” said Reed, “combining a focus on new technologies that are grounded in the needs of the most demanding computational science applications.”
Sterling said that the CCT at LSU is “an exceptional environment” within which to continue his research. He believes the vision, stable funding, and collaborative environment will help bolster his key projects. He is the second of two CCT hires within the computer science department this year.
“Sterling’s presence in the department will bring a new dimension to the reputation and prestige of the department and provides a solid base in terms of research to the department in many areas of high-performance computing,” said Sitharama Iyengar, chair of the computer science department.
“Something new and special is happening at LSU,” said Sterling. “There are experts in-house working with outside people coming in. The old model was for centers of excellence; now it’s changing to circles of excellence. CCT exemplifies this new research methodology.”
CCT Director Ed Seidel said, “We are thrilled to have someone as distinguished and energetic as Thomas joining LSU. The CCT now covers the bases with leading researchers in areas from advanced grid and supercomputing software toolkits to the architectures themselves.” Sterling plans to utilize resources like Access Grid technology for video conferencing and a visitors program, which is available at the CCT and at most top research centers. He believes that these tools support the collaborative model of a circle of excellence.
He is also interested in using the state’s new high-bandwidth optical network. “My team will be supported by the LONI network to facilitate collaboration with other institutions,” said Sterling. The researcher says that a network like this can help to achieve the collective expertise needed to attack the complex problems of science.
Although Louisiana has not traditionally been associated with research in high-performance computing, Sterling was attracted to LSU. The rich research environment and the community impressed him.
“It’s not too rural or too intense,” said Sterling. “I found intelligent people who understand where they are and where they are going. It’s actually a highly cosmopolitan group of people that you would find at any good university.”
Sterling believes that the state’s support of technology at universities is the right model to speed Louisiana forward. “The legislated state funding is based on a local environment with an interest in technology. This is how you bring in talent and leap frog toward success.” Sterling says that he would like to help Gov. Blanco achieve her mission for the state. “I would be proud to be a part of the state’s rise and contribute to it achieving its long-term vision. It can be a driver for the future of the nation.”
The Center for Computation & Technology, or CCT, at LSU is an interdisciplinary research environment for advancing computational sciences, technologies and the disciplines they touch. The center’s efforts serve Louisiana through international collaboration, promoting progress in leading edge and revolutionary technologies in academia and industry. The center is funded by the Louisiana Legislature’s Information Technology Initiative.
Look for an upcoming HPCwire interview with Mr. Sterling!