Sojourn-time analysis on nodal congestion in broadband networks and its impact on QoS specifications

In this paper, we study the sojourn-time statistics and the temporal behavior of nodal congestion in integrated broadband networks. The node is modeled as a finite quasi-birth-death (QBD) process with level-dependent transitions. By formulating the problem as one which is amenable to the generalized folding algorithm (GFA), we are, for the first time, able to analyze realistic systems with large buffer and complex input traffic. The dynamics of the system under realistic traffic environment and different operating regimes are studied. The effects of various modeling artifacts such as fluid-flow and infinite-buffer approximations are also investigated. The potential of statistical multiplexing in reducing bursty cell loss is demonstrated. The trade-offs between different system design alternatives, e.g., buffering vs. statistical multiplexing, are discussed. We also investigate the controlling effect of preemptive cell discarding on steady-state and transient system performance. Both single-level and two-level overload control with hysteresial-switching mechanisms are considered. The use of sojourn-time based QoS metrics to supplement long-term steady-state metrics is also discussed.