Abstract de la publi numéro 17306

Deriving the complete distribution of the end-to-end delay in a wireless multi-hop network is of paramount interest when delay-sensitive flows have to be conveyed over such networks. First works have proposed models to derive the total delay distribution of networks assuming the well-known IEEE802.11 DCF medium access (MAC) protocol. Current derivations can be decomposed into two main steps: (i) the calculation of the total delay probability generating function (PGF) and (ii) its numerical inversion. We show in this paper that there is a need for a thorough performance evaluation of these models since both steps introduce errors, naming modeling and inversion errors. We argue that both types of errors have to be analyzed separately to characterize the accuracy of the analytical derivations of the literature. Therefore, this paper defines two performance evaluation metrics that measure the magnitude of both types of errors. Both metrics are illustrated to select and optimize the most accurate model to calculate the single-hop end-to-end delay distribution of nodes using the IEEE802.11 DCF MAC protocol. The most accurate model is extended to calculate the end-to-end delay distribution for a 2-hop wireless communication.