A layered infrastructure for mobility-aware best connectivity in the heterogeneous wireless internet

The common availability of wireless devices with multiple communication interfaces, e.g., IEEE 802.11, WiMAX, Bluetooth, and/or UMTS, is pushing towards the necessity of novel supports to seamlessly select the proper connectivity technology to exploit at any time. That selection should be context-dependent and consider several aspects, at very different abstraction layers, from application-specific bandwidth requirements to expected client mobility, from connectivity costs and energy consumption to user preferences. We claim the need of effective mobility-aware middleware solutions to relieve application logic from the burden of determining the most suitable interface and connectivity provider for each client at runtime. In particular, we claim that such middleware supports should be structured according to a two-layer architecture: a lower-layer facility to retrieve available interfaces and connectivity providers and to discard unsuitable ones with a per-node decision, e.g., to reduce power consumption; and a higher- layer facility to select the currently most suitable connectivity provider in a per-application way. The paper describes the design and implementation of our novel middleware built according to those architecture guidelines: that permits to clearly differentiate lower-level wireless interface management and connectivity evaluation from higher-level monitoring/selection, thus simplifying the separation between node- and application-specific requirements and the dynamic introduction of new connectivity evaluation metrics. In addition, to take mobility-aware connectivity decisions, our middleware effectively exploits the predicted degree of client node mobility, estimated in a completely autonomous decentralized way. The reported experimental results demonstrate the feasibility of the approach, with accurate estimations of node mobility associated with very limited overhead.