A client-oriented distribution architecture for Web search agents

Among the agents deployed on the Web, search agents which can query multiple search engines simultaneously are gaining popularity among end-users. As the number of users and search engines increase, however, establishing multiple connections to remote search engines has the potential to saturate the network and increase the load on the server running the search agent. To resolve these issues, we propose a client-oriented distribution architecture for Web search agents. Our architecture allows having multiply replicated agents distributed throughout the network initiate the queries while localizing the network traffic to the client's machine. Special attention has been made on the design and implementation of a search agent based on our proposed architecture that limits its use of network resources and prevents skewing the logs on the remote sites. We also present the evaluation results of our Web search agent.