In the field of Networking and Telecommunications, researchers have often the choice between several environments (real, experimental, emulated or simulated) to validate their results. Each facet of our work, from the study of network behavior, to protocol improvements, including the characterization of specific phenomena, can be related to different environments more or less realistic. We are thus confronted to a trade-off mixing the level of abstraction, bias of measurement, and implementation complexity (in term of cost and time). Nevertheless, because simulations are often regarded as quicker, easier to use and simpler environments than real testbeds, they are mostly used despite their accuracy limits. This is particularly true when simulating wireless communication, where cross-layer effects and physical medium are complex to isolate and understand.While some effort have been made to improve simulators, their comparison with real environments are still in progress and requires experimental equipments, especially for wireless networks. In this context, we built a controlled test-bed environment to calibrate wireless simulators. This paper compares the results obtained for scenarii respectively played within this environment and the Ns3 platform. Our analysis method is based on network traces comparison using classification trees. On a first basis the construction of the tree is supervised, but later, it could be unsupervised and feed the need for a comparison tool of different network environments.
