A smart fairness mechanism for Concurrent multipath transfer in SCTP over wireless multi-hop networks

详细信息    查看全文

• 作者：Nasim Arianpoo ; ^{nasima@ece.ubc.ca}Author Vitae ; Victor C.M. Leung ^{vleung@ece.ubc.ca}Author Vitae
• 关键词：Fairness ; CMT-SCTP ; Machine learning ; Wireless ; Multi-hop ; Smart distributed algorithm
• 刊名：Ad Hoc Networks
• 出版年：2017
• 出版时间：February 2017
• 年：2017
• 卷：55
• 期：Complete
• 页码：40-49
• 全文大小：1538 K
• 卷排序：55

文摘

The emerging use of multihomed devices in wireless multi-hop networks has increased the demand for multipath transport protocols, such as Concurrent multipath transfer in Stream control transmission protocol (CMT-SCTP). The fairness of CMT-SCTP over wireless multi-hop settings is a critical issue since it can harm the performance of the existing legacy protocols such as Transmission control protocol (TCP) and single-homed SCTP. As such, in this paper, we study the fairness behavior of CMT-SCTP on a multi-hop wireless testbed. The investigation reveals that CMT-SCTP is significantly unfair towards flows coming from farther away hops. Consequently, we introduce a distributed Q-learning mechanism to enhance the fairness of CMT-SCTP association towards other flows. The proposed method uses Reinforcement learning (RL) to acquire knowledge about network dynamics. The acquired knowledge is used to choose the best action to improve the fairness index of the network. We compare our proposal with standard CMT-SCTP and Resource pool CMT-SCTP (CMT/RP-SCTP). Extensive experimentation confirms that our proposal significantly outperforms the available fairness mechanisms for CMT-SCTP.