Performance evaluation of collaboration algorithms in an agent-based architecture for information retrieval

This paper presents an architecture based on mobile agent technology for retrieving prices of a specific item among many distributed merchant sites. We introduce two collaboration strategies that we implemented and tested: communication through blackboard and inter-agent communication. Empirical performance study was made, based on measurements of response time and induced network load, in the best and worst case scenarios. We also tested the scalability of the system with respect to the number of involved agents. Results show that collaboration between agents gives better performance than non-collaboration and each collaboration strategy is recommended for particular specifications.