基于带宽与往返时间联合预测的多路径并行传输性能优化算法

在流控传输协议( stream control transmission protocol, SCTP)中,多路径并行传输利用多家乡特性实现数据在关联的多条端到端路径中的并行传输。然而,受不同路径性能差异的影响,多路径并行传输将带来接收端的数据乱序。为了减轻数据乱序的程度并提高网络吞吐量性能,需要尽可能准确地估计每条路径的实时带宽与往返时间( round trip time, RTT)。本文利用扩展矢量卡尔曼滤波对多路径并行传输中每条路径的可用带宽与往返时间进行联合预测,同时提出了一种综合考虑发送端未经接收端确认的数据的路径选择算法。仿真结果表明,通过实时准确地预测可用带宽和往返时间,路径选择算法能够减轻接收端数据乱序的程度。对于带宽敏感的多路径应用场景而言,该算法的收敛速度比Kalman-CMT算法更快,对网络吞吐量性能也有一定程度地提高；对时延和带宽都敏感的多路径应用场景来说,算法在收敛速度与吞吐量两方面优势明显。

CMT performance optimization algorithm based on union prediction of bandwidth and round trip time

Concurrent multipath transfer ( CMT) uses the stream control transmission protocol’ s ( SCTP) multihoming feature to distribute data across multiple end-to-end paths in a multihomed SCTP association. Due to the disparity of multipaths, it is facing a great challenge to solve the disorder of received data packets. In order to lighten the reordering degree and then to improve the through-put performance, we need to estimate the bandwidth and round trip time ( RTT) of the real-time paths as exactly as possible. In this paper, we use the extended vector Kalman filter to predict the available bandwidth and RTT of each path simultaneously. Based on this, we propose a predictive path selection algorithm for CMT in SCTP. Simulation results show that the path selection algorithm can lessen the data packets disordering by correctly predicting each path’ s bandwidth and RTT in real time. To bandwidth sensitive scene, the algorithm can converge more quickly than Kalman-CMT and can improve the system total throughput in a certain extent. To time and bandwidth sensitive scene, the algorithm can greatly improve the convergence speed and total throughput than Kalman-CMT.