Cooperative Fault Tolerant Distributed Computing
Graham E. Fagg · 2006
HARNESS was proposed as a system that combined the best of emerging technologies found in current distributed computing research and commercial products into a very flexible, dynamically adaptable framework that could be used by applications to allow them to evolve and better handle their execution environment. The HARNESS system was designed using the considerable experience from previous projects such as PVM, MPI, IceT and Cumulvs. As such, the system was designed to avoid any of the common problems found with using these current systems, such as no single point of failure, ability to survive machine, node and software failures. Additional features included improved inter-component connectivity, with full support for dynamic down loading of addition components at run-time thus reducing the stress on application developers to build in all the libraries they need in advance.