Wireless networking outside of the simulator

Summary form only given. Simulation is an important tool in the development of mobile ad hoc networks. It has a number of important benefits in providing an environment to experiment and verify routing protocol correctness. However, there is no guarantee that the protocol works in practice because simulators contain assumptions and simplified models that may not reflect real network operation. For instance, while most widely used mobility models create a dynamic environment that can be used to stress a protocol in a mobile network, they do not generate realistic movement patterns. It has been shown that mobility model selection directly influences the performance of the routing protocol. Hence, the use of an unrealistic model results in predictions that do not reflect achievable performance in the real world. Other sources of predication inaccuracy stem from the omission of relevant, up-to-date protocol models. As a result of these simplifications, inaccuracies and omissions, mobile network performance can be incorrectly predicted. In particular, the performance of field networks can be orders of magnitude worse than the simulated performance of the same network. This poses a serious problem for mobile networking researchers, who have invested heavily in simulation as a tool for network development and evaluation. The talk describes some of the disparities between predicted and measured performance, and discuss mechanisms for bridging this gap.