Single Server Queueing Models for Communication Systems

One of the most important performance measures in the analysis of data networks is the average delay required to deliver a packet from origin to destination. The packet delay will aaect the choice and performance of several network algorithms such as routing and ow control. Queueing theory is the primary methodological framework for analyzing network delay. It provides a basis for adequate delay approximations, as well as valuable qualitative results and worthwhile insights. The literature on queueing models for communication systems is abundant. In this paper we concentrate on single server queueing models for communication systems. We describe some of the more recent models that were developed for modeling the behavior of nodes of modern communication networks. These models are fundamental and provide the foundation for more elaborate models with multiple servers. We focus on the queueing and transmission delays of packets. First, we describe some classical queueing models, namely, the M=M=1, M=G=1 and G=M=1 queueing models. These models assume that all interarrival times and service times are independent. This assumption does not hold for communication systems and we describe extensions of the classical queueing models to allow for such dependency. It is demonstrated that with this dependency the delays in the buuers preceding communication links are smaller. Next, we describe completely diierent approaches for the characterization of arrival processes. These characterizations are motivated by the need to model real-time applications in high-speed communication networks where it is important to capture the burstiness of the arrival processes. Finally, we describe another performance measure, namely, the message delay. The motivation for this measure arises from the process of segmentation and reassembly of messages into packets which is natural for packet-switched networks.