Fault Tolerant PVM Application ? — A Case Study with D - CICADA ( FT )

The ever increasing number of computers connected into network leads to the development of new programming techniques to allow full use of the potential of these virtual computers. Despite problems of programming difficulty, associated with the distributed computing environment, the problem of fault tolerance is also emerging. The distributed computing environment is often used as a low-cost substitute of fast (super)computers. However, the very nature of these virtual computers, created from dozens of ordinary workstations in local or metropolitan area networks, is not really suitable for long time tasks. Congestion in the network, software and hardware crashes, broken connections etc are all the reasons while the efficient use of the virtual computer is further complicated. To ease its use, modified and/or new tools and techniques must be provided, both at the level of distributed computing environment and at the level of programming methodology. In this article, D-CICADA(FT), a robust implementation of the D-CICADA system [1] is presented. It is shown how careful design may lead to robust (and almost fault tolerant) system.