From advanced, high-performance clustered systems to highly scalable, shared memory and vector systems, SGI solutions uniquely serve the specialized computing requirements across technical markets. The 512-processor SGI 2800 server installed at NASA Ames is the largest shared memory system in the SGI Origin 2000 Series. It provides researchers with a single system image, allowing for easier system administration and software program development while at the same time providing unprecedented performance. Within the first week of operation the SGI 2800 system was performing production calculations with the OVERFLOW Computational Fluid Dynamic code at a sustained rate in excess of 50 Gflop/s per second.
"With the SGI 2800 system, a team of engineers can evaluate the performance of aircraft or space vehicle concepts early in the design process. The end result is that better design decisions can be made early in the process, ultimately leading to much better vehicles at lower cost," said McDonald.
According to McDonald, in the past, aerospace designers could make only limited use of highly accurate, but computationally intensive, simulations early in the design process. Therefore critical early design decisions could not be made with the best possible information.
Building a fast, 512-processor shared memory system is a requirement for complex modeling applications, such as computational fluid dynamics, where the interactions of a variety of individual elements are being simultaneously analyzed. With the power of the new system, designers can routinely use more accurate design tools, leading to better design data and better design decisions. For example, with the new system, a highly complex 35-million point computational fluid dynamics aircraft simulation can be completed in a few hours rather than days or weeks.