Performability analysis of wireless mobile communication systems

Components of wireless mobile systems are subject to unplanned and planned outages. The duration of the outages and recoveries may follow arbitrarily general distribution in contrast to the widely-held exponential assumption for convenient Markov availability models in practice. We study the availability of systems with non-exponentially distributed outages and give useful criteria with regard to the use of all-exponential availability models. We also study how redundancy in a system can be taken advantage in planned outages to improve system performability.; We introduce a fault-tolerant feature to mitigate the adverse impact by control channel failures. We build Markov chain performability models to show the effectiveness of the automatic protection scheme in the cellular environment with multiple base repeaters for each cell. To conduct efficient numerical evaluation, we also build hierarchical Markov models using decomposition technique to avoid largeness and stiffness in the exact models.; The advent of packet-switched data services into wireless mobile systems presents great challenges in the performance modeling. One of the major difficulties is that the arrival processes of connection requests are so complicated by the data user behavior and the strong coupling effect caused by the multiple access mechanisms that simple assumptions such as Poisson arrivals may no longer be valid. We take on this problem by first examining the performance of a realistic system, GPRS (General Packet Radio Service), by building a comprehensive model which reflects the major features in its uplink multiple access mechanism. We then investigate a downlink scheduling policy, namely the Generalized Weighted Queue Length (GWQL) policy, to evaluate its performance under bursty data traffic and time-varying channels. We also study the effect of the distribution of access delay in multiplexing data traffic over a wireless link under the capacity-on-demand concept in wireless data services. (Abstract shortened by UMI.)...