An Experimental Study on Energy Consumption of Wireless Multipath TCP Connections

Multipath TCP (MPTCP) allows multihomed devices to exploit the available bandwidths of multiple radio access technologies (multi-RATs) of overlapped coverage. Moreover, it provides the necessary data transfer reliability and congestion control. However, the energy consumption behavior of different MPCTP congestion control algorithms remains unclear. This paper offers a comparative experimental study for the energy consumption performance of three MPTCP congestion control algorithms, namely, Cubic, linked increases algorithm (LIA), and opportunistic linked increases algorithm (OLIA). The lab experimentation is performed using a testbed that contains dual-homed wireless devices, which send some amount of data via MPTCP over two wireless interfaces. One is connected to a real (not simulated) LTE eNodeB and the other to a real WiFi access point (AP). The three algorithms are compared under varying factors such as data flow lifetime, receiver buffer size, and data size. Our findings reveal that the three algorithms consume similar amount of energy with long-lived flows, whereas OLIA exhibits the best energy consumption behavior for relatively short-lived flows. Also, it is observed that OLIA consumes the lowest amount of energy if an intra-eNodeB and/or AP handover occurs during the lifetime of a long-lived flow.