Enhancement of reliable data transport performance with 1-hop TCP connection relay in wireless mesh networks

• Main Authors: Hsuang-En Wei, 魏祥恩
• Other Authors: 林啟芳
• Format: Others
• Language: zh-TW
• Online Access: http://ndltd.ncl.edu.tw/handle/31097348946423015573

碩士 === 元智大學 === 資訊工程學系 === 100 === The throughput and reliability of TCP suffer when it is used in mesh networks. This is because TCP needs end-to-end Acknowledgement to finish every transmission, so single error will trigger the original sender to retransmission data no matter where error happened. This Mechanism resulting overall degradation of throughput especially affects connections with a large number of hops, where error rate is more likely; thus, short connections enjoy more advantage. We develop a scheme, called single-ack mechanism, which split long TCP connection into single-hop connections. For every node in this scheme, receiving data and immediately return acknowledgement to the sender and terminate the transmission, then relay data to destination with same mechanism. This scheme lost TCP reliability when a node crash after it send an acknowledgement to sender. To solve this problem we develop a new scheme, called double-ack mechanism, which delay the transmission termination timing. We terminate connection after data relayed two hops away from sender. To achieve this we add one more acknowledgement and sent it to sender after relay data to another node. We test double-ack mechanism in real world environment, and emulate wireless bit error by linux firewall iptable to random drop packet. Result show that double-ack mechanism had smaller effect to the number of hop.