Evaluating the Impact of Packet Scheduling and Congestion Control Algorithms on MPTCP Performance over Heterogeneous Networks
Dimitrios Dimopoulos, Apostolis K. Salkintzis, Dimitris Tsolkas, Nikos Passas, Lazaros Merakos
TL;DR
This study analyzes how MPTCP performance is shaped by packet scheduling and congestion control in heterogeneous networks. It combines a thorough theoretical treatment of MPTCP architecture with an extensive Mininet-based experimental campaign that evaluates 5 schedulers (minRTT, BLEST, ECF, RR, LLHD) across 7 CCAs (CUBIC, LIA, OLIA, BALIA, wVegas, BBR, C-MPBBR) under 29 heterogeneity scenarios, introducing PS_score and CCA_score metrics. The results show no universally optimal pairing; however, C-MPBBR, BBR, and OLIA consistently deliver strong performance across scenarios, with minRTT/ECF performing well under several CCAs. The work provides actionable guidance for configuring MPTCP in mixed networks and establishes a framework for comparing future MPTCP algorithm combinations, highlighting the trade-offs between throughput, delay, and fairness in real-world conditions.
Abstract
Modern mobile and stationary devices are equipped with multiple network interfaces aiming to provide wireless and wireline connectivity either in a local LAN or the Internet. Multipath TCP (MPTCP) protocol has been developed on top of legacy TCP to allow the simultaneous use of multiple network paths in the communication route between two end-systems. Although the combination of multiple paths is beneficial in case of links with similar network characteristics, MPTCP performance is challenged as heterogeneity among the used paths increases. This work provides an overview of the MPTCP protocol operation, analyzes the state-of-art packet scheduling and congestion control algorithms available in literature, and examines the impact of the various algorithm combinations on MPTCP performance, by conducting an extensive experimental evaluation under diverse path-heterogeneity conditions.