| Computing systems are increasingly distributed in nature and information is no longer confined to a single computer, but stored on various computers and can be under constant change. As a result traditional computing paradigms, such as the client-server communication model used for Web communications, are increasingly insufficient since they are unable to scale to these environments. New computing models will require a certain degree of autonomy that allows rapid decisions based on current, and likely local, information in order to achieve desired, long-range objectives. Mobile software agent systems have several key characteristics that are well suited for these challenging computing environments.;Many mobile agent systems rely on a population of agents roaming a network performing various tasks. The number of agents in the population is often critical for ensuring certain visitation rates throughout the network. However, it is not uncommon for the number of agents to change, due to issues within the network, such as component failure. This thesis establishes a scheme for managing the population of dynamic mobile systems by governing the death and birth of agents through the examination of the expected visitation rates of randomly wandering agents. Furthermore to avoid oscillation of the population of agents (complete death and rebirth of all agents) and visitation rates, agent queue management is analyzed. Finally this thesis performs an empirical analysis of a death and birth algorithm for agents and a queue management algorithm, that seek to prove network monitoring scalability and resilience. |
