On the performance of traffic-aware reactive routing in MANETs

Al-Qassas, Raad S (2007) On the performance of traffic-aware reactive routing in MANETs. PhD thesis, University of Glasgow.

Abstract

Research on mobile ad hoc networks (MANETs) has intensified over recent years, motivated by advances in wireless technology and also by the range of potential applications that might be realised with such infrastructure-less networks. Much work has been devoted to developing reactive routing algorithms for MANETs which generally try to find the shortest path from source to destination. However, this approach can lead to some nodes being loaded much more than others in the network. As resources, such as node power and channel bandwidth, are often at a premium in MANETs, it is important to optimise their usage as far as possible. Incorporating traffic aware techniques into routing protocols in order to distribute load among the network nodes would helps to ensure fair utilisation of nodes' resources, and prevent the creation of congested regions in the network. A number of such traffic aware techniques have been proposed. These can be classified into two main categories, namely end-to-end and on-the-spot, based on the method of establishing and maintaining routes between source and destination. In the first category, end-to-end information is collected along the path with intermediate nodes participating in building routes by adding information about their current load status. However the decision as to which path to select is taken at one of the endpoints. In the second category, the collected information does not have to be passed to an endpoint to make a path selection decision as intermediate nodes can do this job. Consequently, the decision of selecting a path is made locally, generally by intermediate nodes. Existing end-to-end traffic aware techniques use some estimation of the traffic load. For instance, in the traffic density technique, this estimation is based on the status of the MAC layer interface queue, whereas in the degree of nodal activity technique it is based on the number of active flows transiting a node. To date, there has been no performance study that evaluates and compares the relative performance merits of these approaches and, in the first part of this research, we conduct such a comparative study of the traffic density and nodal activity approaches under a variety of network configurations and traffic conditions. The results reveal that each technique has performance advantages under some working environments. However, when the background traffic increases significand, the degree of nodal activity technique demonstrates clear superiority over traffic density. In the second part of this research, we develop and evaluate a new traffic aware technique, referred to here as load density, that can overcome the limitations of the existing techniques. In order to make a good estimation of the load, it may not be sufficient to capture only the number of active paths as in the degree of nodal activity technique or estimate the number of packets at the interface queue over a short period of time as in the traffic density technique. This is due to the lack of accuracy in measuring the real traffic load experienced by the nodes in the network, since these estimations represent only the current traffic, and as a result it might not be sufficient to represent the load experienced by the node over time which has consumed part of its battery and thus reduced its operational lifetime. The new technique attempts to obtain a more accurate picture of traffic by using a combination of the packet length history at the node and the averaged number of packets waiting at node's interface queue. The rationale behind using packets sizes rather than just the number of packets is that it provides a more precise estimation of the volume of traffic forwarded by a given node. Our performance evaluation shows that the new technique makes better decisions than existing ones in route selection as it preferentially selects less frequently used nodes, which indeed improves throughput and end-to-end delay, and distributes load more, while maintaining a low routing overhead. In the final part of this thesis, we conduct a comparative performance study between the end-to-end and on-the-spot approaches to traffic aware routing. To this end, our new load density technique has been adapted to suggest a new "on-the-spot" traffic aware technique. The adaptation is intended to ensure that the comparison between the two approaches is fair and realistic. Our study shows that in most realistic traffic and network scenarios, the end-to-end performs better than the local approach. The analysis also reveals that relying on local decisions might not be always good especially if all the potential paths to a destination pass through nodes with an overload condition in which case an optimal selection of a path may not be feasible. In contrast, there is most often a chance in the end-to-end approach to select the path with lower load.

• Item Type: Thesis (PhD)
• Qualification Level: Doctoral
• Additional Information: Adviser: M Ould-Khaoua
• Keywords: Computer science
• Date of Award: 2007
• Depositing User: Enlighten Team
• Unique ID: glathesis:2007-73993
• Copyright: Copyright of this thesis is held by the author.
• Date Deposited: 23 Sep 2019 15:33
• Last Modified: 23 Sep 2019 15:33
• URI: https://theses.gla.ac.uk/id/eprint/73993