Priority queueing systems and its applications to congestion control

Abstract

In order to support services with different quality of service requirements, various priority controls have been proposed in packet networks such as ATM networks and Signaling system No.7 networks. When congestion occurs in a packet network, proper congestion control should be triggered and the packets should be treated differently according to their priorities. In this thesis, we analyze priority queueing systems and we consider their applications to the performance analysis of Signaling system No.7 networks. In Chapter 3, we analyze a priority queueing system with congestion control based on thresholds. A triple of thresholds consisting of an abatement threshold, and onset threshold and a discard threshold was considered by San-qi Li. We extend Li``s method to $N$ triples of thresholds to control packets according to priorities, and obtain the equilibrium probabilities of the system by using the Generalized Folding Algorithm and the Embedded Markov chain method. We obtain the loss probability and the mean delay of each class packet. We apply our result to signaling system No.7 network with a congestion control based on thresholds. In Chapter 4, we analyze the transient behavior of a queueing system with congestion control based on thresholds. We assume that the arrival processes are Markovian arrival processes(MAP) which are generalization of Poisson processes and Markov modulated Poisson processes. By considering embedding Markov chain at arrival epochs, we derive the one step transition probabilities. We obtain the mean delay and the loss probabilty of packets at arrival epochs. We apply our result to signaling system No. 7 network with a congestion control based on thresholds. In Chapter 5, we analyze a leaky bucket control scheme in the signalling system No.7 network. Burst of correlated messages can be sent through the SS7 network if access is available to user signalling bearer capabilities as specified in the CCITT Q.931 standard. These user-to-user...