A framework for the evaluation of routing protocols in opportunistic networks

Abstract

The evaluation of routing protocols for opportunistic networks can be seen as a multidimensional problem because it involves several performance aspects. To capture these aspects various evaluation metrics are used, such as the number of delivered packets, the delivery delay and the number of transmissions. Unfortunately, in the context of opportunistic networks, these metrics are often highly correlated and usually conflicting. To make things worse, the characteristics of the network affect the importance of each metric as well as the levels of its correlation with other metrics. In this work, we first propose a set of performance evaluation metrics that are normalized with respect to the optimal performance. This approach tackles several of the above-mentioned shortcomings of traditional evaluation metrics. We then formulate the evaluation of routing protocols as a Multiple-Criteria Decision-Making (MCDM) problem where each routing protocol is an alternative and the performance metrics correspond to a set of criteria. We use this formulation to develop an evaluation framework that objectively ranks the performance of opportunistic routing protocols. To this end, we reshape well-known concepts and algorithms from the MCDM field to address the special requirements that are specific to the opportunistic context. We present detailed simulation results of well-known routing protocols in various opportunistic environments and rank their performance according to the proposed framework. In conclusion, no algorithm was able to achieve the best performance in all or the majority of the network topologies that we studied. This demonstrates the diversity of challenges that routing mechanisms face in such networks.