Mobile Agents for Flow Scheduling Support in Multihomed Mobile Hosts

Mobile host multihoming provides benefits to both users and network providers. A mobile terminal's (MT) capability of simultaneously serving user traffic through multiple radio interfaces enhances the performance and reliability of wireless access. On the other hand, the availability of alternative paths for reaching end-hosts gives network operators the opportunity for load balancing their radio access networks. This work focuses on efficient utilization of available network resources from the perspective of a multi-mode MT, multihomed through its multiple radio interfaces. In the presence of device, application and user specific constraints, efficient utilization involves: i) selection of the host's radio interfaces that must be active, ii) selection of the best access network to be associated with each active radio interface, iii) assignment of application flows to the most appropriate radio interface in a way that maximizes user utility. For brevity reasons, the term flow scheduling is used in this work to refer to the above decision problem. We show that flow scheduling maps to a binary integer programming problem that is generally hard to solve for large problem sizes. In order to support efficient flow scheduling, in the usually resource constrained MTs, an agent- based architecture is proposed that delegates flow scheduling functionality to mobile agents (user representatives) executing in the network side. Performance implications of this approach related to the latency introduced to the decision process, due to agent messaging, are studied through simulations and the results are promising for the feasibility of our approach.