An Experiential Study of Multipath TCP Schedulers With the Comparison of Throughput, Latency, and Path Utilization

Abstract

Advanced internet technology and data-hungry applications encourage us to use Multipath TCP (MPTCP). MPTCP can use many homogenous or heterogeneous interfaces like LTE, Wi-Fi, Ethernet, etc. simultaneously. This inherently introduces out of order packets(OFO) scenario at the receiver side. If packets are not collected in sequence at the receiver then after a timeout, retransmission and spurious retransmission may occur thus increasing overhead and decreasing the quality of user experience. To optimize this, an efficient MPTCP scheduling algorithm becomes essential. This article studies and compares a few MPTCP schedulers theoretically and practically using key performance metrics. Comparative study of data points suggests that known schedulers of MPTCP do perform well in uniform network environments, but are inflexible to adapt to dynamic environments with the constant flux in network states of wired/wireless and homogeneous/heterogeneous setup. This calls for a flexible algorithm that can learn the network state and adapt to it for optimized network performance.