The application of intelligent software agents in network fault diagnosis

Abstract

With the advent of existing and future networking, the task of network fault monitoring and in particular that of fault diagnosis is becoming quite complex. Agents can be used to help in this domain. A software agent is a piece of software that is autonomous, goal-oriented and temporally continuous. Agents can be built on different architectures including reactive and deliberative. Agents can communicate with each other using languages such as FIPA’s ACL or KQML. Most network management applications currently us SNMP. SNMP is reliable but has a low degree of flexibility and is centralised. It was designed for less complex types of networks than those that currently exist. When comparing traditional network management applications with an agent-based applications, one needs to consider utilisation as well as fault diagnosis. Network utilisation is concerned with how a network is being used, and how much and what type of traffic is travelling through network segments. When designing an agent-based network management system, it must be fast enough to react quickly to changes in the network. It should also minimise the amount of interaction that is needed with the network manager.