Very high performance networking for supercomputing

NASA Ames has installed a very-high-bandwidth, 1 Gb/s local-area network (using two UltraNet 1000 hubs) to study the feasibility and performance of networking supercomputers with minisupercomputers and workstations at the effective bandwidth needed for computing environments of the future. A network display device, which provides interactive graphics over the network, was also installed to study the impact of this new bandwidth on visualization and the impact of interactive graphics on network performance. Three classes of computers were defined for study: workstations, minisupercomputers, and supercomputers. Interactions between these classes in a typical operational environment were also studied. The author presents the results of investigations of network performance in a production facility through distributed processing.<
>