Thorough Empirical Analysis of X-MAC Over a Large Scale Internet of Things Testbed

The last decade saw the emergence of the Internet of Things (IoT) paradigm, which aims to connect any object to the Internet. As in IoT networks, the radio component is usually the main source of energy depletion, specific medium access control protocols have been designed to reduce its usage. Among those, X-MAC has gained much interest in recent years, and has become one of the most popular MAC protocols in the IoT community due to its many theoretical advantages. X-MAC is a preamble-sampling protocol based on the low power listening mechanism, in which the nodes decide their schedule independently from their neighbors. So far, no study has ever identified malfunctions that appear in the protocol design. In addition, performance evaluations were only made through theoretical studies and simulations, or relatively limited scale experiments. In this paper, we point out several omissions and black spots that appear in the original protocol design and that lead to encountering severe underachievements. We also provide explanations to these flaws and propose solutions that solve or at least reduce their impact. Finally, we have led a thorough experimental campaign on an indoor static grid composed of 240 nodes. From those studies, we quantify the performances of the original X-MAC design and the impact of our improvements.