Collaborative Environments for Scientific Computing The Task of Algorithm/Software Selection
Naren Ramakrishnan, Anupam Joshi, Elias N. Houstis, John R. Rice · Purdue e-Pubs (Purdue University System) · 1997
The advent of High Performance Computing (HPe) and the Internet has brought forth a 'net-centric' scenario for scientific computation where important pieces of algorithms and software arc spread across a(thc) network.The currenL networked computing scenario assumes that the choice of method to solve a scientific problem is fLx:ed a priori and that appropriate code, modules are downloaded, compiled and linked to form static programs.However, lhe selection of an appropriate method is by no means obviolls and suitable methodologies do not exist to aid in the mapping of problem solution requirements to appropriate algorithms/sofl.ware.Moreover, the network-based paradigm also allows for a scenario wherein multiple servers can house appropriate algorithms; this brings up the need for collaborative systems that dynamically select among competing servers/repositories and advise about algorithms/software.In this paper, we present a methodology that performs automatic algorithm selection for well-defined domains in scientific computing.Our approach selects among competing servers to determine the 'best' resources based on a notion of 'reasonableness'.This is implemented in the context of distributed I