High-Level System Specification with iSPNs

In the previous chapter, QBD processes have been introduced as a powerful class of stochastic processes which is amenable to very efficient solution algorithms. Furthermore, the QBD framework provides a concise and memory-efficient description of the infinitely large generator matrix. Thus, from a computational point of view, it is desirable to employ QBD processes as a low-level mechanism for the numerical analysis. From a user perspective, however, the development of QBD models directly at the Markov chain level suffers from the same drawbacks as the direct specification of large finite Markov chain models: the model development is cumbersome and error-prone, as is the derivation of the performance measures of interest.