GAMING

To add significant computing and storage capacity to existing life sciences research infrastructure, the Bioinformatics Consortium at the University of Missouri recently purchased high-performance computing technology from Silicon Graphics and an SGI InfiniteStorage storage area network (SAN) with 8TB of capacity. The new SGI Altix server was specifically chosen for its shared-memory architecture and Linux operating environment, which will allow researchers, particularly in areas of molecular dynamics, molecular modeling, protein structure prediction, and a number of allied activities, to run larger-scale problems than ever before. Installed in November, the SGI systems were made available to university researchers beginning in mid-December. The Altix will immediately take over a large number of big computational chemistry jobs currently being run on a slower server, including structural studies of bipolymers, chemistry in interstellar space, and polar order in crystalline organic molecular materials. Researchers studying environmental issues using satellite and GIS data, and other scientists studying electrical properties of solid-state devices, have already requested blocks of time on the SGI Altix system as well. There is also very early-stage research of electronic structure, crystal structures and chemical structures of various biological compounds with the idea of being able to use natural organic compounds to develop computational capabilities. The primary reason the University purchased the SGI Altix system is to run these kinds of problems, which require massive processing power and shared-memory architecture to get results in hours instead of days or weeks. "One of the key inhibiting factors for researchers who want to run larger problems is the size of memory," said Dr. Gordon Springer, Scientific Director, Bioinformatics Consortium, University of Missouri. "With the SGI Altix we're significantly increasing our capacity for large-scale computations requiring large memory, problems that are even beyond the scope of our existing system, which is hardly small. We've run some benchmarks, and jobs that were taking about 50 to 60 hours on the HP Alpha SC system actually took about 2 to 6 hours on a similarly configured SGI Altix. I consider the Altix to be a keystone in our infrastructure. The Altix was, I believe, the most appropriate system that we could put in place that would support the symmetric multiprocessing large-scale shared-memory environment for life sciences researchers across the university system." Life sciences are a strategic area for the University of Missouri, which encompasses four campuses: Kansas City, St. Louis, Rolla, and the flagship campus in Columbia. The University of Missouri Bioinformatics Consortium was founded in July 2001, with the mission of providing infrastructure to support the life sciences researchers across the University system. The University carries out joint research projects in medicine, and in life sciences generally, and maintains active relationships with Missouri-based medical IT firms, agricultural corporations such as Danforth Plant Sciences Center, biotech companies including Monsanto, and the Stowers Institute for Medical Research in Kansas City. The Bioinformatics Consortium at University of Missouri purchased an SGI Altix 3700 Bx2, with 128GB RAM and 64 Intel Itanium 2 processors running Novell SUSE Linux Enterprise Server with the SGI ProPack software performance suite, and an 8TB SGI InfiniteStorage SAN with InfiniteStorage shared filesystem CXFS to support instant data access for large analysis, simulation or modeling programs. The main applications the SGI Altix system will run are Gaussian, Amber 8, and NAMD, popular open-source tools for chemistry, biochemistry and physics, especially for studying molecular and crystal structures. The Bioinformatics Consortium now has three different architectures to support the life sciences initiative. As Dr. Springer noted, the SGI Altix system fills the needs of large memory SMP architectural kinds of problems while a Linux cluster based on dual-core Intel Xeon processors will support some of the more readily available, but not quite as highly parallelized, kinds of programming applications that many researchers want to run. A third FPGA (field programmable gate array) system is focused on implementing bioinformatics algorithms for large scale data comparisons and genomic database searching. "Given the complement of different systems and different architectures, we hope to be able to capture and be able to support effectively all of the life sciences researchers, regardless of the specific kinds of computational needs that they require," said Dr. Springer. "In fact, the Xeon system and the Altix will be sharing the same SGI TP9500 storage system, which has 8TB, and the two systems will be talking on the same set of disk. I do distributed computing and grid computing and so, through some of the code that we'll be developing, I expect to make this environment seamless for the researchers. We would actually schedule jobs based upon the architecture that best suits their computing needs." The Bioinformatics Consortium resources are networked university-wide and to the Abilene 10Gb Internet2 network nationally. The University of Missouri is looking into building a high-speed network for inter-campus communication as well as looking at alternatives, including joining the National LambdaRail, a dark fiber high-speed network rapidly being adopted by leading universities nationwide. The Bioinformatics Consortium also supports some projects on animal and plant sciences, including the possibility of engineering animals for organ transplants for humans. Four years ago, University of Missouri genetic engineering researchers were the first to clone a pig using the techniques for transgenic nuclear transfer. Subsequent experiments have identified two genes responsible for material in a pig's heart that causes instantaneous rejection in humans. The University has now grown a herd of pigs without those two genes, meaning that the possibility exists—by inserting human DNA into a pig—to actually customize a pig heart for a human transplant with little or no immune system rejection, perhaps in five to 10 years. Dr. Springer expects some data analysis work will be done on the SGI Altix system, as researchers start moving into trying to understand the chemistry and molecular structure of some of these genes. "The addition of the processing power of the SGI Altix to the University of Missouri's Bioinformatics Consortium will encourage scientists from many disciplines to move from their lab desktops into the world of large datasets, which can be run in hours not days, or weeks," said Afshad Mistri, senior manager of scientific markets, SGI. "The results of that research will, in time, benefit everyone, and the training will enable students to easily move into jobs in biotech firms, government research labs and medical IT systems companies where the shared-memory architecture and power of SGI systems are being rapidly adopted."