ACADEMIA
Harvard Implements Largest IBM Blue Gene Supercomputer in US Academia
The CrimsonGridBGL will enhance multi-scale computational science and support advanced research in science and engineering: IBM, in conjunction with Harvard University’s Division of Engineering and Applied Sciences (DEAS), announced the implementation of CrimsonGridBGL, the largest IBM System Blue Gene Solution in United States academia. The system would currently rank among the top 50 fastest supercomputers in the world, according to the latest statistics gathered from the TOP500 Supercomputer Sites list.
The CrimsonGridBGL, offering a peak performance of 11 trillion floating point calculations per second (teraflops), will help faculty and researchers explore multi-scale computational science and applications across a variety of disciplines. This announcement marks an extension of IBM’s prior collaborations—in particular, the establishment of the Crimson Grid, a computing grid established in 2003 for faculty and student research, data sharing, and collaboration in a variety of areas, including life sciences, engineering, and applied sciences. “Increasingly sophisticated computational tools and mutually-reinforcing industry collaborations are instrumental to the success and advancement of research across all areas,” said Venkatesh Narayanamurti, dean of Engineering and Applied Sciences at Harvard. “The new system comes at an ideal time for us and compliments the University’s plans to dramatically increase its investments in science and engineering.” The initial projects involving the Blue Gene platform are likely to be dedicated to modeling complex, vast systems or events such as: the human hemodynamic (blood circulation) system; cell self-assembly and tissue morphogenesis as they relate to fundamental processes underlying cardiac organogenesis (the development of the heart); computer system behavior; the mechanical response of materials used in advanced integrated circuits; and the formation history of galaxies. “With the arrival of the Blue Gene system, we are thrilled that we will be able to work even more closely with researchers and students at Harvard,” said David Turek, vice president of Deep Computing at IBM. “We look forward to our continued collaboration with the goal of seamlessly combining traditional clusters and Ultra-scale systems like the Blue Gene system as our next step.” The Blue Gene system packs 4,096 PowerPC processors into two racks and takes up less than three-square meters. Relative to the equivalent amount of processors in a traditional cluster, the system uses four times less space and consumes five times less power. To gain parallel scalability the Blue Gene system uses a standard MPI foundation; this allows complex problems to be calculated simultaneously among its thousands of processors and low-latency connections. “With the tremendous computational capabilities of the Blue Gene system, research deployment, or the ability to handle multiple projects at the same time, will increase five-fold,” said Jayanta Sircar, CIO at DEAS and the Director of the Crimson Grid Project. “The existing grid infrastructure, which you can think of as an entry point to accessing the Blue Gene system, will provide a consistent and integrated high speed network for managing workflow.” Sircar expects the CrimsonGridBGL will become a workhorse not only for computationally complex problems in science and engineering but for broader areas—from financial risk analysis to epidemiology—that span the entire University.