From gophers to ants-a case for mobile agent families

Mobile agents can be used to minimize network traffic for distributed applications through the execution of specialized selection and concentration functions in the location of remote information providers. So far, this optimization potential has only been studied in small-scale scenarios. In view of an envisioned application field in a global electronic information space, the issue of scalability to very large-scale scenarios has to be addressed. Analysis results presented indicate that the popular multi-hop agent approach does not scale in communication costs for large sets of information providers. This paper proposes a new approach based on families of one-hop mobile agents which is shown to scale, and discusses design issues for the required runtime support. Additional motivations for the one-hop schema are discovered in the fields of group communication protocols, network monitoring and agent security.