|
Abstract : |
This paper presents a status report on our efforts and experiences with a metacomputing project that aims at connecting a Cray T3E at Pittsburgh Supercomputing Center and a Cray T3E at the High Performance Computing Center in Stuttgart. We describe how the two machines are configured into a single virtual machine running code with standard MPI calls. The library that handles the underlying communication management is presented. This library provides the user with a distributed MPI environment with most of the important functionality of standard MPI. The first application we have run on the coupled T3E's is the flow simulation package URANUS. Problems and strategies with respect to metacomputing are discussed briefly for this code. First results both for the library performance and the code performance are given for tests on both local machines and machines connected via vBNS. These tests will serve as a starting point for further development. Investigation of these results shows that latency will be the critical factor for all applications except those that are embarrassingly parallel. KEYWORDS:, |
