Article Preview
Top1. Introduction
With the rapid growth of Internet of Things (IoT) and mobile communications, the need for QoS guarantees has become of primary importance, especially when hand-over events occur by Mobile Hosts (MHs) changing coverage areas during their active sessions; they may find scarce resource availability in new locations and the current active connections can be dropped. To the best of our knowledge, the only way to ensure QoS and service continuity to mobile users consists of making a bandwidth reservation over all the cells that a MH will visit during its active connection. There are many protocols able to ensure early reservations like Next Step In Signaling (NSIS) (Fu, 2005), Dynamic ReSerVation Protocol (DSRVP) (Huang, 2004) and Mobile ReSerVation Protocol (MRSVP) (De Rango, 2009), but a prediction scheme is mandatory in order to know which coverage cells a user will probably visit during its Call Life Time (CLT). On the basis of previous works (De Rango, 2005; Fazio, 2012), we considered the MRSVP, which gives the possibility to exchange the right communication messages among the predicted coverage cells, achieving the needed passive amount of bandwidth in the cells where the MH will probably hand-in. The same Markov model has been considered, but an optimization on the number of chain states is now proposed: in the previous contributions, only one hand-over direction has been considered for the hand-off event toward a next cell, without considering the roads topology that characterize MH movements. Given that the number of chain states could be very large if all the roads that lead to another cell are considered, an optimization scheme is proposed. In particular, the dynamic programming approach is considered (Shivaram, 1997), having the possibility to choose the right number of states for the Markov model, taking into account the morphology of the considered geographical region. An approximation has been introduced and the associated error has been minimized. Clearly, in order to implement and realize this kind of prediction, a real network operator has to analyse users’ mobility, through a statistical treatment. In our case, without access to real data about MH movements, we employed the Citymob for Roadmaps (C4R) mobility generator (Martinez, 2008), in order to appreciate prediction performance when mobility traces are extracted from real roadmaps of different countries (the mobility model has a heavy impact on the obtained results, that may be unsuitable if the adopted mobility model is unrealistic). The integration between the Markov process and the dynamic programming approach leads to a new distributed prediction scheme, called Dynamic Markov Prediction Algorithm (DMPA), tested through extensive simulation studies. The rest of the paper is organized as follows: section 2 gives an overview of the existing related work, section 3 gives a detailed description of the proposed scheme, by considering the environment and the solution. Section 4 shows our simulation results, then section 5 concludes the paper.