Abstract
Wireless network research in the last years is often based on simulation. Ns-2 is a widely used wireless network simulation tool for this purpose. However, there are no published results about the accuracy of the ns-2 wireless model in the literature so far. In this paper we present the validation of one wireless network model built with ns-2 done by comparing the network characteristics of a simulated, an emulated, and a real wireless network. In order to show only the relevant differences, we have calibrated the radio propagation model of ns-2 to the real network and have used the same routing protocol implementation and the same application data traffic in all the compared networks. The results show that the packet delivery ratios, the connectivity graphs, and the packet latencies are represented in the model with an average error of 0.3%, 10%, and 57% respectively. Based on these results we conclude that the packet delivery ratios, and network topologies are accurately represented in ns-2, once the simulation parameters are properly adjusted. The accuracy of the packet latencies is lower and therefore statements about latencies in the real network based on the simulation results have a lower validity. Based on these results we provide recommendations for future development of the ns-2.
1 Introduction
Wireless network simulation has been a widely used method to develop and evaluate wireless networks in the last years. Its advantages are that the network scenarios are easily constructed and modified, the models are easily monitored from the global view of the simulator, and the experiments are reproducible. Testing conditions that are hard to achieve in reality are easily constructed in simulation. A recent review on wireless network research papers [1] from an ACM symposium based on 151 articles from a five-year-period reported that 76% from the works used network simulation. This fact shows the wide usage of simulation in wireless network research.
Among the tools used for wireless network simulation, the network simulator ns-2 [2] is one of the most often used. In the review cited above the ns-2 holds the first place with 44% usage. Research using ns-2 is also reported in other IEEE/ACM journals and conferences. Applications and network protocols are being developed at different network layers and for different types of systems for instance resource allocation [3], real-time communication [4], and energy issues [5] in ad-hoc networks, transport protocols in wireless sensor networks [6],
and control strategies for wireless robots [7]. The keywords ”wireless”, ”mobile”, or “ad-hoc” can be seen in about 30% of the messages posted to the ns− users mailing list within the time period Feb.–Sept. 2006. These facts show that ns-2 is an actively used network simulation tool for research and development in the area of wireless networks.
Some authors [8] state that ns-2 has inaccuracies in modelling the physical layer of wireless networks. Although we agree with these authors, it is also true that no model is 100% accurate [9]. The needed accuracy of the model depends on the objectives of the simulation study. However, to the best of our knowledge, no experience on the accuracy of a wireless network model build with ns-2 has been published in the literature so far (see section 2). Therefore, up to now it is hardly possible to make a decision whether to use ns-2 or not in a simulation study with some specific demand on the model accuracy. In this paper we fill this knowledge-gap and provide experimental results on the accuracy of an ns-2 wireless model for different network performances. We achieve this by doing a validation of a typical wireless model build with ns-2. Validation is the process of comparing a model and the corresponding system based on some particular points of interest [9]. In our study these are different network performances: network connectivity graphs, packet delivery ratios, and latencies.
Regarding the principles of discrete-event simulation [9], every simulation model goes through a calibration and validation phase to achieve the needed accuracy. Therefore, the results from our model validation are not directly applicable to every model build with ns-2. However, as we are using a typical mesh network, it is highly possible that the accuracy of other typical mesh network models are close to what we have achieved. Furthermore, the experimental methods and tools used for validation in this paper are general and can be used for other validation studies. Therefore, the scientific contributions of this work are:
1) the experimental results for the accuracy of one ns-2 wireless model and
2) the experimental methods and tools for validation of network models.
The paper is structured in the following manner: Section 2 provides an overview on the related work. Section 3 describes the validation methodology and the experimental methods used for the validation. In section 4 we provide experimental results that show the differences between simulation, emulation, and a real wireless network. Finally, we provide recommendations for future development to bring the ns-2 and its wireless model closer to reality.
Svilen Ivanov, Andr´e Herms, and Georg Lukas
Institute for Distributed Systems
Otto-von-Guericke University of Magdeburg
Universit¨atsplatz 2
D-39106 Magdeburg, Germany
{svilen,aherms,glukas}@ivs.cs.uni-magdeburg.de
No comments:
Post a Comment