A new approach to the analysis of Markovian queueing networks is developed and applied to traffic problems of voice-data integration networks and trunked mobile radio networks. This approach is shown to be computationally much less complex for large systems compared to previous work.
In the integrated networks we consider, two classes of users share the system facilities. If all the servers are busy, the first class of users are queued, but the second class of users are blocked and cleared from the system. The performance objective of such an integrated network is to trade the time delay performance of the first class of users against the blockage performance of the second class of users to keep the grade of service as high as possible for both classes of traffic.
The key-state approach introduced in this thesis is what makes the analysis of the corresponding Markovian queueing network model of these integrated networks computationally less complex than that of previous work. The performance of integrated networks is investigated under several control strategies and new exact closed-form expressions are obtained for the equilibrium probabilities of the corresponding Markovian models. The results are extended to a more general Markovian process where a bulk of arrivals and departures are allowed.
The key-state approach is expected to become a standard tool for analyzing large queueing networks such as will arise when Integrated Services Digital Networks (ISDN) become widely deployed in the next five years.
