Adding value to WSN simulation using the IoT-LAB experimental platform

Validation of protocols and mechanisms is an essential step to the development of object networks in critical domains. Most papers still provide evaluation either obtained through theoretical analysis or simulations campaigns. Yet, simulators and formal models fail to precisely reproduce the unique specificities of the deployment environments those networks have to evolve in. Also, by putting no limits to code complexity and execution, those tools prevent users to apprehend the actual limits of WSN nodes and to propose realistic communication protocols and applications. In this paper, we highlight to what extent the addition of experimentations can significantly improve the value of performance evaluation campaigns. Along with the recent tendency to have algorithmic and protocol proposals facing real environments, it is questionable whether the so obtained results should be considered as scientific or empirical ones. In the former case, reproducibility, stability over time, topology management to cite a few, are a must have for testbeds and real deployments that are used. In the latter, the results should be viewed as a proof-of-concept only, far from independent of the used hardware and encountered conditions at the experimentation time but still critical from the development cycle standpoint. Through some experiments over the IoT-LAB testbed, we aim at demonstrating to what extent some of the simulation setup and conditions from reality could be emulated. We also provide insight on how to obtain the best out of it in a quick and efficient manner. We show that such testbeds would satisfy many expectations (e.g. scientific tool and proof-of-concept validator), thus minding and bridging some of the gaps between theory and practice in WSN. To this end, we here give an overview of available simulation tools, and guidelines on how to transpose simulation setups to the open large-scale IoT-LAB platform.