Energy-efficient scheduler for MPTCP data transfer with independent and coupled channels

Over the last few years, one can observe a shift in preferences to employ wireless media, including WiFi and cellular, rather than wired technologies for the communication purposes. Since even simple networking devices are nowadays equipped with a few communication interfaces, the user's experience can be elevated by simultaneous transmission over different paths, as offered by the recently standardized Multipath TCP (MPTCP). This paper addresses the problem of optimal load distribution among the interfaces available at a networking device, e.g., a mobile phone, with the objective of minimizing the overall energy expenditure. Depending whether the corresponding data paths are unrelated or influence each other, two cases are considered and two dynamic allocation algorithms are designed. The design procedures are set on the formal grounds of convex optimization and extremum seeking control. An experimental comparison with the earlier MPTCP scheduling solutions shows superiority of the proposed algorithms in terms of energy requirements. In particular, it is shown that in the case of coupled channels, the energy savings can reach several percent without throughput degradation.