The increasing availability of mobile devices with wireless communication capabilities stimulates considering new service
provisioning environments where accessibility to the traditional Internet is provided via Access Points (APs) working as bridges
between fixed hosts and wireless devices. The most notable example is the case of IEEE 802.11 APs that support connectivity
of Wi-Fi terminals to a wired local area network. In the following, we will indicate these integrated networks with fixed
Internet hosts, wireless terminals, and wireless APs in between, as the Wireless Internet (WI).
WI opens new challenging scenarios for mobile service provisioning. On the one hand, time-continuous services, such as audio/video
streaming, relevantly suffer from the temporary disconnections that mobile clients experience at their handovers from old Wi-Fi
access localities to new ones; in addition, they often require moving reached session states in newly visited localities. On the
other hand, WI pushes towards the possibility to provide novel services whose contents depend on client location; location dependency
complicates application design and implementation, and requires innovative support functions.
We claim that both mobile services with session continuity requirements and location-dependent ones can relevantly benefit from the
adoption of lightweight and decentralized mechanisms capable of
predicting wireless client handover between WI
access localities. In particular, the idea is to provide support functions for
- handover prediction, to notify interested services that one of their clients is going to change AP in a given time interval,
- mobility prediction, to notify interested services that one client is going to leave its current access locality by specifying which
is the most probable next AP.
For instance, handover prediction
could help continuous services to pre-fetch data chunks just before handovers, thus simplifying service interruption avoidance while
preserving wireless client memory when handovers are not probable. Mobility prediction can permit to perform location-dependent
service management operations, e.g., resource re-bindings, in next WI localities before the actual connections of mobile clients.
Our work on handover and mobility prediction is part of our original middleware for the support of WI service provisioning, which
transparently mediates wireless client access to distributed applications via mobile proxies that dynamically adapt service results
to client terminal properties, location, and runtime resource availability.
Let us stress that our prediction solution is completely local and decentralized: each wireless client hosts its own predictor,
whose results only depend on either actual or filtered RSSI values for all APs in visibility, with no need of interacting with
either other clients in its WI access localities or support components running in the wired infrastructure. Our middleware simply
exploits the local RSSI monitoring data that IEEE 802.11 client cards have to collect anyway to be compliant with the standard; the
middleware-level awareness of RSSI data is achieved in a completely portable way over heterogeneous platforms, as detailed in.