Supercomputing '95: Computing and Visualizing Einstein's
Gravitational Waves across the Metacenter
Supercomputing '95: Computing and Visualizing Einstein's
Gravitational Waves across the Metacenter For Supercomputing '95, the NCSA group is developing a distributed, heterogeneous, scientific application that will solve the full 3D Einstein equations for the gravitational field, while simultaneously exploiting the resources at various sites around the metacenter and displaying the results as they are computed in the CAVE. This application is a continuation of work shown previously at both Supercomputing '93 and SIGGRAPH '94. At those conferences we demonstrated both the usefulness of the CAVE for scientific visualization and the feasibility of using high speed networks to distribute a complex scientific application across local HiPPI or FDDI networks. For Supercomputing '95 we plan to explore the use of the I-WAY ATM network to distribute the calculation across the largest machines available in the metacenter (thereby permitting the largest simulation of Einstein's equations ever attempted), and also to refine our scientific visualization capabilities in the CAVE.Our group, which is part of the NSF funded Grand Challenge effort to solve the Einstein equations, has already transformed one our 3D codes (H3expresso) to begin this effort. Using message passing (through MPI, PVM, or other libraries) the code can already run efficiently on the CM-5, SP-1, T3D, and SGI Power Challenge, and it has already been tested on a heterogeneous computing environment linking NCSA and Argonne. The Einstein equations offer an excellent opportunity to illustrate heterogeneous, scalable computing, as very large amounts of computation must be performed at each point in the computational domain, leading to a very high computation to communications ratio needed for success in such an application. Based on our experience to date, and the resources available at the metacenter, we believe that such a distributed computation should be able to achieve speeds of order 100 GFlops. This effort will provide an important testbed for development of other large scale distributed scientific applications, as our numerical algorithms are common to many scientific and engineering disciplines, therefore having direct application to other fields of computational science.
For this demonstration, we will require resources at a number of centers, such as the NCSA CM-5, the Cornell SP-2, the Pittsburgh C-90 and T3D, and the San Diego Paragon. We have already requested time on all these machines through the MAC Grand Challenge grants this year. We are considering additional sites, such Los Alamos, Argonne, and others. We will also require the CAVE, which will be used both to show the results of the computations and to control the distributed code application while it is running.