Abstract
The experience with the development and application of a distributed problem-solving paradigm over local ATM networks is reported in this article. The paradigm consists of a collection of problem-solving agents that autonomously execute and coordinately cooperate each other to achieve goal. It is implemented based on Java Expert System Shell (Jess) and SUN's Remote Procedure Call (RPC). However, communication between agents has been the bottleneck of the paradigm. The most likely path uses emerging commercial technologies such as ATM networks serving the communication need. The architecture and performance issues for local ATM networks meeting the paradigm requirement are described. We then treat the real application in network fault management and compare the performance of three different interfaces including DEC's ATM API, RPC, and Parallel Virtual Machine (PVM). The results show that the paradigm can obtain a high degree of goodness in management applications over local ATM networks.