SIMULATION SOLUTIONS OF NETWORKS WITH PRIORITISED TRAFFIC INVOLVING THE MULTI-PRIORITY DUAL QUEUE

摘要

In prior work we have shown that the multi-priority dual queue (MPDQ) outperforms conventional scheduling disciplines such as First In First Out in isolation. This work takes the MPDQ into a network situation, and compares the network loss with and without its presence. As the MPDQ improves traffic congestion, most notably in communication networks, it is not necessary to include it at every node, or service centre, within a network. Our aim here is to provide a framework for future researchers, network designers, and service providers to implement this analysis, which involves the investigation of simple three, four and five node multi-class networks. This is aimed as a guide for extension to larger networks. Queueing networks containing differentiated traffic, also known as multi-class networks, are complicated to solve analytically using existing queueing theory techniques. We discuss the complications of exact probability network solutions and describe how we used involved simulation models to obtain performance statistics only possible through high performance computing. The simulation models here are presented with a description of the workings of the MPDQ. To investigate if there is any improvement in loss levels within the network we compare, for each network, the inclusion of one or more MPDQ's. Each node contains a service centre, following specific service times for each traffic type. Traffic arrives and follows a shortest path algorithm to its predetermined destination.