[论文解读] TCP Congestion Control Scheme for Wireless Networks based on TCP Reserved Field and SNR Ratio

Youssef Bassil|arXiv (Cornell University)|Jul 4, 2012

Network Traffic and Congestion Control参考文献 18被引用 4

一句话总结

本文提出了一种新型的无线网络TCP拥塞控制方案，利用TCP保留字段将链路分类为有线或无线，并使用信噪比（SNR）来区分拥塞引起的丢包与误码引起的丢包。通过根据链路可靠性调整拥塞窗口行为，该方案在不降低有线网络性能的前提下，提升了误码率较高的无线环境下的吞吐量。

ABSTRACT

Currently, TCP is the most popular and widely used network transmission protocol. In actual fact, about 90% of connections on the internet use TCP to communicate. Through several upgrades and improvements, TCP became well optimized for the very reliable wired networks. As a result, TCP considers all packet timeouts in wired networks as due to network congestion and not to bit errors. However, with networking becoming more heterogeneous, providing wired as well as wireless topologies, TCP suffers from performance degradation over error-prone wireless links as it has no mechanism to differentiate error losses from congestion losses. It therefore considers all packet losses as due to congestion and consequently reduces the burst of packet, diminishing at the same time the network throughput. This paper proposes a new TCP congestion control scheme appropriate for wireless as well as wired networks and is capable of distinguishing congestion losses from error losses. The proposed scheme is based on using the reserved field of the TCP header to indicate whether the established connection is over a wired or a wireless link. Additionally, the proposed scheme leverages the SNR ratio to detect the reliability of the link and decide whether to reduce packet burst or retransmit a timed-out packet. Experiments conducted, revealed that the proposed scheme proved to behave correctly in situations where timeouts were due to error and not to congestion. Future work can improve upon the proposed scheme so much so that it can leverage CRC and HEC errors so as to better determine the cause of transmission timeouts in wireless networks.

研究动机与目标

• 解决TCP在无线网络中因无法区分拥塞引起的丢包与误码引起的丢包而导致的性能下降问题。
• 通过基于链路特性的拥塞控制自适应机制，提升异构有线-无线网络中的TCP效率。
• 利用现有的TCP首部字段（保留字段）和SNR测量值，实现智能丢包分类，而无需修改协议。
• 在保持与标准TCP向后兼容的同时，提升在不可靠无线链路上的性能。
• 基于实时链路质量动态调整传输行为，避免不必要的窗口减小。

提出的方法

• 利用TCP保留字段来指示连接是否运行在有线或无线链路上。
• 在接收端监测信噪比（SNR）以评估链路可靠性并检测误码率较高的状况。
• 根据链路类型应用不同的拥塞控制逻辑：有线链路采用标准TCP行为，无线链路采用修改后的行为。
• 当发生超时后，检查SNR以判断丢包是由于误码（低SNR）还是拥塞（高SNR）引起。
• 若丢包归因于误码（低SNR），该方案会触发重传但不减小拥塞窗口。
• 若丢包归因于拥塞（高SNR），则按标准TCP行为减小拥塞窗口。

实验结果

研究问题

• RQ1 TCP保留字段是否可被有效重用于指示链路类型（有线与无线），而不会破坏向后兼容性？
• RQ2 SNR测量在多大程度上能可靠地区分无线网络中拥塞引起的丢包与误码引起的丢包？
• RQ3 与标准TCP相比，该方案在混合有线-无线环境中如何提升TCP吞吐量与公平性？
• RQ4 该方案是否能在误码引起的超时期间避免不必要的拥塞窗口减小，同时保持稳定性？
• RQ5 通过SNR进行链路质量估计对无线场景下TCP的整体可靠性与响应能力有何影响？

主要发现

• 所提出的方案成功利用SNR和保留字段区分了误码引起的丢包与拥塞引起的丢包。
• 在误码率较高的无线环境中，该方案在由比特错误引起的超时期间避免减小拥塞窗口，从而保持了吞吐量。
• 实验结果表明，与标准TCP相比，该方案在无线环境中实现了更高的吞吐量和更短的重传延迟。
• 该方案通过使用保留字段而非修改核心协议语义，保持了与现有TCP实现的向后兼容性。
• 利用SNR作为可靠性度量可实现对链路退化情况的准确检测，从而对传输问题做出适当响应。
• 未来改进可集成CRC和HEC误码检测，以进一步细化无线网络中丢包原因的识别。