Stochastic analysis of upper delay bound of GPS-based packetized fair queueing algorithms

Fair queueing (FQ) algorithms which aim at approximating the generalized processor sharing (GPS) policy remain most popular for the provision of quality-of-service guarantees in IP networks. In this paper, we extend the notion of feasible partition introduced by Zhang et al., for the analysis of idealized GPS discipline and apply it to the stochastic bound analysis of GPS-based packetized FQ algorithms. A novel upper bound on mean packet delay is derived under the probabilistic traffic model of Poisson arrival and any general packet length distribution and it is much simpler and tighter than the known ones by M. Hawa et al., (2002). Moreover, the derived upper bound fits a class of GPS-based packetized FQ algorithms including WFQ, SCFQ and SPFQ.