This project investigated the design of scientific applications for multi-tier network systems using distributed object technology. A multi-tier system is one with several levels of data locality and process parallelism, for example, a cluster of SMP nodes, or a cluster of distributed-memory parallel systems. Many factors needed to be considered:

  • the performance impact of different network node hardware, which could range from single processor workstations to shared-memory or distributed-memory multiprocessor systems;
  • the different interconnects that provide communication between nodes, ranging from slow Ethernet to high-performance, custom-designed interconnect for multiprocessor systems;
  • the portability, performance and security of the underlying system software used by the application.

As a concrete example of the impact of these factors on scientific applications we developed a distributed version of a parallel multiphysics application called PHYSICA.

Here are some of the reports.

MetaPhysica: theory and reality
Status report on distributed PHYSICA implementation.
Tim Oliver. February 1999.
(Email me for a copy of this document.)
PHYSICA: A distributed multiphysics application
Distributed PHYSICA design.
Tim Oliver. August 1998.
(Email me for a copy of this document.)
Distributed objects
This short note presents background information about distributed object technologies.
Tim Oliver. March 1998.
(Email me for a copy of this document.)

Last updated May 26, 2004
Copyright © Tim Oliver. All rights reserved.