The OptIPuter: A New Approach To Volume Visualization Of Large Seismic Datasets
Graham Kent, John Orcutt, Larry Smarr, Jason Leigh, Atul Nayak, Debbie Kilb, L. Renambot, S. Venkataraman, Tom DeFanti, Y. Fialko, Phil Papadopoulos, Greg Hidley, David Hutches, Maxine Brown, Ocean Technology Conference, May 2004.
Introduction
Scientists at Scripps Institution of Oceanography, in collaboration with researchers at UCSD, the Electronic Visualization Laboratory [EVL] at the University of Illinois at Chicago [UIC], and the San Diego Supercomputer Center [SDSC] have embarked on a project to redefine how visualization is optimized at the hardware, software, and networking layers. We anticipate this will dramatically improve rendering performance, while significantly reducing cost. The premise behind this new approach to visualization computing is the hollowing-out of the computer. In this innovative approach, individual parts of a computer, such as CPUs, memory, graphic-boards, disk drives, are interconnected via the network rather than backplanes or busses. This is now possible because of the on-going development of fast networks and switches that enable 10 Gigabit (OC-192) speeds or greater. As network speeds increase, less importance is placed on the location of the individual parts of a computer, but on the network layer that connects all the various pieces, and the software that is optimized to run in this environment.