Fault Diagnosis for Local Area Network Environments

Particular attention has been given to propose systems to solve network management tasks, especially for the fault diagnosis and performance management. The construction of such systems requires an intense process of modeling the network environment through interviews with experts in the network management area as well as the use of Artificial Intelligence techniques. The aim of this paper is to specify Problem Solving Methods for the diagnosis of communication network faults. We claim that the AI approach called Model-based diagnosis provides a foundation for exchanging behavioral, structural and control information between the subtasks of such complex systems. We also show what are the main aspects to be considered when constructing such systems, namely: how to build network models (manually or automatically); how to model an appropriate problem solving method to each class of network faults; how to identify the type of interaction between the diagnosis system and a network status gathering system, such as a management platform; how to construct the communication interfaces among several systems, etc. Finally, this work presents two prototypes of diagnosis systems: one for configuration faults and another for communication faults of conventional TCP/IP local area networks. *