Improving the Fairness of Distributed Queue Dual Bus (DQDB) Network

Abstract

This thesis is concerned with the study of improving the fairness of Distributed Queue Dual Bus (DQDB) network. The Distributed Queue Dual Bus (DQDB) protocol is selected by the IEEE 802.6 working group as the preferred access method for Metropolitan Area Network (MAN). However, many studies have reported that large propagation delay to transmission time ratios (high bit rates, large networks) can cause a certain degree of unfairness in access delay which depends on the stations position with respect to the head ofbus station (slot generator). More importantly, this unfairness behavior, which depends on the size of the subnetwork, can degenerate into a deadlock (hogging) situation where one station can monopolize almost all the available bandwidth. The distribution ofbandwidth during an overload period is a function of the initial network conditions and is in that sense unpredictable. This problem was first illustrated in [16,17] with a number of hogging scenarios. In these scenarios, one node can take and keep almost 100% of the bus bandwidth regardless of other nodes demands, given that it has a head start. To solve this problem, an enhancement to the DQDB protocol, the Bandwidth Balancing Mechanism [19], was proposed and adapted by the IEEE 802.6 working group. This scheme limits the maximum bandwidth available to each node in order to leave a portion of bandwidth unused. This unused bandwidth permits the congestion to relax and allows equal sharing ofthe available bandwidth to be attained eventually. Unfortunately, further studies [22] found that the Bandwidth Balancing Mechanism is ineffective with multiple priorities. We study the fundamental reasons of this unpredictable behavior by means of the bandwidth distribution equations of DQDB with BWB in overload conditions and we provides graph model which meets these criterion of DQDB with BWB and validate the steady state of the DQDB(BWB) network under heavy load conditions with different initial conditions (i.e., the bandwidth of the network is equally distributed among the all active nodes) by using simulator. The thesis consists of six chapters, a conclusion, and a comprehensive list of references. After the introduction, which provides the brief description of the thesis. Chapter 1 is presented to initiate the reader to take a brief history about the computer networks.

Chapter two, explains the Topology, and the perfonnance of the QPSX (Queued Packet and Synchronous Switch) that is the first version ofDQDB has been provided for a public Metropolitan Area Network. Chapter three, introduces more details about the Architecture ofDQDB Network i.e., Dual Bus Structures and Looped Bus Structure and explains also the interaction between Data Link Layer (DQDB layer) with the Physical Layer in the MAN Network. In forth Chapter, the DQDB Protocol and its perfonnance under overload traffic conditions is presented, also the original DQDB access mechanism for a synchronous traffic is reviewed with more details and the unfairness problem is clearly analyzed by using analytical analysis [17]. Chapter five, presents the concept of Bandwidth Balancing (BWB) mechanism, and the

implementation of BWB mechanism using the modified DQDB method, also the 11 Graph Model 11 , which is based on the space-time diagram, for the case of two active (overloaded) nodes in the network, which completely specifies the nodal throughputs. In chapter six, the simulation results are obtained for various possible scenarios in the case of two active nodes in the network, and these results satisfied the analysis of modified DQDB under steady state condition by using space time diagram for the two station case (heavy loaded) with three different initial conditions. The graph model and the simulator results illustrated clearily the transition of the DQDB with BWB network from transient state to the steady state under different initial conditions and network parameters. There is one Appendix at the end of the thesis containing the listing of the simulator program.