Eigenanalysis solution for quasi birth and death process
The behavior of many systems of practical interest in communications and other areas is well modeled by a single server exponential queueing system in which the arrival and service rates are dependent upon the state of a Markov chain, the dynamics of which are independent of the queue length. Formal solution to such models based on Neuts's matrix geometric approach have appeared frequently in the literature. A major problem in using the matrix geometric approach is the computation of the rate matrix, which requires the solution of a matrix polynomial. In particular, computational times appear to be unpredictable and excessive for many problems of practical interest. Alternative techniques which employ eigenanalysis have been developed. These techniques are polynomially bounded and yield results very quickly compared to iterative routines. On the other hand, the class of systems to which the eigenanalysis based techniques apply have been somewhat restricted. In this thesis, we modify the eigenanalysis approach initially presented in order to remove some of these restrictions.