Topics on modelling and simulation of wireless networking protocols

Jeffrey W. Wildman II

doi:10.17918/etd-3009

The use of computer simulation to study complex systems has grown significantly over the past several decades. This is especially true with regard to computer networks, where simulation has become a widespread tool used in academic, commercial and military applications. Computer model representations of communication protocol stacks are used to replicate and predict the behavior of real world counterparts to solve a variety of problems. The performance of simulators, measured in both accuracy of results and run time, is a constant concern to simulation users. The running time for high fidelity simulation of large-scale mobile ad hoc networks can be prohibitively high. The execution time of propagation effects calculations for a single transmission alone can grow unmanageable to account for all potential receivers. Discrete event simulators can also suffer from excessive generation and processing of events, both due to network size and model complexity. In this thesis, three levels of abstracting the Institute of Electrical and Electronics Engineers (IEEE) 802.11 Request to Send/Clear to Send (RTS/CTS) channel access mechanism are presented. In the process of assessing the abstractions' ability to mitigate runtime cost while retaining comparable results to that of a commercially available simulator, OPNET, the abstractions were found to be better suited to collecting one metric over another. Performance issues aside, simulation is an ideal choice for use in prototyping and developing protocols. The costs of simulation are orders of magnitude smaller than that of network testbeds, especially after factoring in the logistics, maintenance, and space required to test live networks. For instance, Internet Protocol version 6 (IPv6) stateless address autoconfiguration protocols have yet to be convincingly shown to cope with the dynamic, infrastructure-free environment of Mobile Ad hoc Networks (MANETs). This thesis provides a literature survey of autoconfiguration schemes designed for MANETs, with particular focus on a stateless autoconfiguration scheme by Jelger and Noel (SECON 2005). The selected scheme provides globally routable IPv6 prefixes to a MANET attached to the Internet via gateways. Using OPNET simulation, the Jelger-Noel scheme is examined with new cluster mobility models, added gateway mobility, and varied network sizes. Performance of the Jelger-Noel scheme, derived from overhead, autoconfiguration time and prefix stability metrics, was found to be highly dependent on network density, and suggested further refinement before deployment. Finally, in cases where a network testbed is used to test protocols, it is still advantageous to run simulations in parallel. While testbeds can help expose design flaws due to code or hardware differences, discrete event simulation environments can offer extensive debugging capabilities andevent control. The two tools provide independent methods of validating the performance of protocols, as well as providing useful feedback on correct protocol implementation and configuration. This thesis presents the Open Shortest Path First (OSPF) routing protocol and its MANET extensions as candidate protocols to test in simulated and emulated MANETs. The measured OSPF overhead from both environments was used as a benchmark to construct equivalent MANET representations and protocol configuration, made particularly challenging due to the wired nature of the emulation testbed. While attempting to duplicate and validate results of a previous OSPF study, limitations of the simulated implementation of OSPF were revealed.

Topics on modelling and simulation of wireless networking protocols

Files and links (1)

Abstract

Metrics

Details

Topics on modelling and simulation of wireless networking protocols

Files and links (1)

Abstract

Metrics

Details

Drexel University Social media