EBA: Efficient Bandwidth Aggregation for Connected Vehicles With MPTCP
Imran Khan, Kang Chen · IEEE Internet of Things Journal · 2021
Multipath TCP (MPTCP) recently emerges as a promising tool to improve the network service to vehicles through transparent aggregation of wireless infrastructures on the Earth (e.g., current terrestrial BSs) and over the air [e.g., unmanned aerial vehicles (UAVs)-based BSs], e.g., using both LTE and WiFi at the same time for a data transfer. Currently, such an approach is challenged by the fast dynamism of wireless link characteristics (e.g., RTT and loss rate) during the handoff between wireless access points under vehicle mobility, which could easily cause Head-of-Line (HoL) blocking at the receiver and significantly lower the overall network performance. The existing solutions generally rely on asynchronous scheduling at the sender to tackle the HoL blocking. However, they fall short in handling the packet losses and excessive delay during the handoff. In this article, we propose a unique MPTCP scheduling scheme named efficient bandwidth aggregation (EBA) to tackle the drawback in current approaches. EBA adaptively enables packet duplication over parallel paths (i.e., redundant scheduling), which can effectively prevent the HoL blocking but losses some bandwidth. Thus, we carefully model the conditions under which packet duplication can improve network performance. We then dynamically turn on and off redundant scheduling based on the modeling to achieve the best bandwidth performance for vehicles. We have implemented EBA inside the current Linux kernel MPTCP and tested it extensively through simulated scenarios. Experimental results show that EBA can effectively enhance the network service for connected vehicles.