TRAFFIC MODELLING IN DISRUPTION-TOLERANT NETWORKS
Rossitza Goleva, Seferin Todorov Mirtchev · 2010
End-to-end Quality of Service management and analyses in DisruptionTolerant Networks requires a dynamic and flexible dimensioning approach. Connection is usually not performed on single path. Multipathing and multihoming in aggressive environment is the only way to transmit reliably traffic bundles. This paper proposes analytical/simulation solution for endto-end performance analyses. The input flows are modeled by Polya distribution that is applicable for continue and discrete time systems. It is also appropriate for peakedness and packet long range dependence simulations. We use flexible queuing bounds for packets’ loss and delay combined with priorities. The bounds depend on cross-layer Quality of Service mapping at application, bundle, transport and IP traffic characteristics. The solution is combined with scheduling techniques at IP level such as Priority Queuing, Weighted Fair Queuing and Round Robin. The results for DiffServ algorithms are shown. The results for single interface with Polya distribution traffic sources and fixed packet length show the necessity of short queues for real time services. When queues with more than 50 places are applied the long range dependence between packets in the queue is seen. Furthermore, we calculate end-to-end delay bounds for real time and non real time services to show how the delay bounds could be used for traffic regulation for delay intolerant traffic. Our calculations demonstrate the capability of the approach for better traffic shaping of real time service on the favor of non real time services.