Multipath TCP Scheduling Optimization Based on PSRBP in Heterogeneous Network

Abstract

Nowadays, with the large-scale deployment of multiple network interfaces in heterogeneous networks, the diversity of interfaces makes the characteristic differences between transmission paths tend to be obvious. However, when a heterogeneous network uses MPTCP for transmission, a large amount of out-of-order data packets accumulate in the receive buffer due to the different characteristics of each path, causing congestion of the receive buffer and resulting in poor MPTCP transmission performance. Aiming at the problem of MPTCP receive buffer blocking, this paper proposes a multipath scheduling optimization scheme named Path Selection and Receive Buffer Prediction (PSRBP)). The scheme predicts throughput based on the throughput prediction model, introduces parameter that indicates the similarity of other paths to the path with the highest throughput, and compares it with the established threshold to select the paths that can be transmitted by MPTCP. Afterward, the proposed scheme estimates the size of the receive buffer by judging the number of out-of-order packets and predicts the buffer blocking in advance and stops the data transmission on the bad paths to improve the performance of MPTCP. The simulation results show that the proposed scheme can improve the overall throughput of the system, making the network utilization rate improved.