INDUSTRY

Wichita State University sciences, math and engineering researchers now have access to the most powerful computing architecture available today with the installation of a SGI Altix 3000 supercluster. More than 130 faculty members, research associates, graduate students and postdoctoral fellows in such fields as computational fluid dynamics, structural and design engineering, and chemistry currently use the system installed in the university's High-Performance Computing Center. Planned research projects using the center's new computing power vary widely, depending on the academic interests of participating faculty members: -- Elizabeth Behrman of the physics faculty is leading a group mapping how quantum computing could one day bring about faster, more powerful computers-up to a million times faster than any currently existing computer -- Robert Carper of the chemistry department is heading a group investigating the molecular structure of ionized liquids, which has important applications for many manufacturing industries -- In a study that reflects the high interest in aerospace engineering and aircraft design issues in the Wichita region, which is an important industry hub, Klaus Hoffmann of the aerospace engineering department is investigating a computational fluid dynamics application showing flow fields under the influence of magnetic fields The new system complements a 24-processor SGI(R) Origin(R) 2000 server that remains in production. As John Matrow, a system administrator/trainer at the center, puts it, "Compared to the previous system, the new processors are seven times faster, and the capacity is four times greater. It's like a highway with a speed limit seven times faster than it used to be and with four times the number of lanes, creating a factor of 28 times the improvement." The system is built on the open source Linux(R) operating system using the latest Intel(R) Itanium(R) 2 processors. A customized feature of the system, installed in July in the university's High-Performance Computing Center, is the integration of the 32-processor Altix system for large-scale, complex computational jobs with two 32-processor Intel Architecture 32-bit clusters for jobs requiring a simple distributed computing model. It's the equivalent of a car with two speeds, one for intensive challenges and the other for everyday driving. SGI Professional Services has integrated yet a third component, a front-end file server dedicated as an easy-to-use interface for the supercluster. Users needn't make decisions about which subsystem will work best for any individual task and can instead concentrate on scientific research tasks. "Most of our users haven't used resources at big supercomputing centers," explains the university's David Alexander, director, High Performance Computing Center, and project manager, Enterprise Resource Planning. "So our design criteria for this mixed environment was a system as transparent and flexible as possible from the users' point of view, and we've accomplished that. SGI did an excellent job of providing an integrated system that's easy for users to work with." "The university's High Performance Computing Center is quickly developing a national profile as a model for successful campus-wide shared computing, as reflected in its ever-increasing government and related funding," says Bill Bartling, Senior Director, Market Strategy, SGI. "It's gratifying to know that the center has reinforced its long-time commitment to an SGI infrastructure with its purchase of this latest system. We very much look forward to working with center staff for many years to come." SGI Professional Services provided the expertise to integrate third-party hardware, software, and services with SGI compute platforms, including an SGI storage component, to create a fully integrated solution that addressed the customer's diverse system requirements. "Most of our users are using third-party software, such as Gaussian," explains Alexander, referring to a widely used software package for computational chemistry. "We're not experts in tuning Gaussian, so having access to SGI engineers who can make sure Gaussian is running effectively is important to us. In fact, an important consideration in our buying decision was the extent to which the vendor could provide service for software we were planning to run." Wild Open Source, Inc., a Linux and Open Source Professional Services company based in Burlington, Mass., worked closely with SGI to implement the cluster solution. "It was great to work with the combined teams from WSU and SGI," remarked Jim Lieb, Wild Open Source's Engineering Consultant. "The overall result was a great architecture that proves the value of SGI products, our combined services, and Linux technology." To learn more about SGI Professional Services, visit http://www.sgi.com/services/, and for information about SGI's involvement with universities and research labs around the world, visit http://www.sgi.com/go/research.