| 引用本文: | 潘婉苏,李晓风,谭海波,许金林,李皙茹.BBR拥塞控制算法的RTT公平性优化[J].哈尔滨工业大学学报,2022,54(11):38.DOI:10.11918/202109034 |
| PAN Wansu,LI Xiaofeng,TAN Haibo,XU Jinlin,LI Xiru.RTT fairness optimization of BBR congestion control algorithm[J].Journal of Harbin Institute of Technology,2022,54(11):38.DOI:10.11918/202109034 |
|
| 摘要: |
| Google提出了一种基于瓶颈带宽和往返传播时间的拥塞控制算法(bottleneck bandwidth and round-trip propagation time,BBR),可以在网络链路中保持最大传输速率和最小延时。然而一些评估实验表明,BBR算法会导致不同往返时间(round trip time,RTT)的数据流之间存在严重的公平性问题。为了优化这一问题,研究分析了BBR算法探测机制所导致的发送速率与瓶颈带宽不匹配对RTT公平性的影响,提出了一种基于起搏增益模型的优化算法BBR-adaptive（BBR-A）。BBR-A算法不再采用原BBR算法中固定的起搏增益,而是利用RTT与起搏增益的关系,构造一个基于反比例函数的起搏增益调节模型,通过让向上和向下的起搏增益系数相互交错来平衡发送速率,使每个BBR流可以公平地竞争带宽资源。网络模拟器3(network simulator 3,NS3)仿真实验结果表明:BBR-A算法的信道利用率比BBR算法有了小幅提升；在RTT公平性的方面,BBR-A缩小了不同RTT流之间的吞吐量差异,在不同缓冲区和RTT差异下,Jain公平指数至少提高了1.5倍；BBR-A算法明显降低了重传率。因此通过自适应调整起搏增益系数,可以平衡不同数据流之间的发送速率,有效提升BBR算法的RTT公平性。 |
| 关键词: 拥塞控制 BBR RTT公平性 起搏增益 自适应算法 |
| DOI：10.11918/202109034 |
| 分类号:TP 393 |
| 文献标识码:A |
| 基金项目:国家重点研发计划“区块链”重点专项(2021YFB2700700) |
|
| RTT fairness optimization of BBR congestion control algorithm |
|
PAN Wansu1,2,LI Xiaofeng1,2,TAN Haibo1,XU Jinlin1,LI Xiru1
|
|
(1.Hefei Institute of Physical Science, Chinese Academy of Sciences, Hefei 230031, China; 2.University of Science and Technology of China, Hefei 230026, China)
|
| Abstract: |
| Google proposed a congestion control algorithm based on bottleneck bandwidth and round-trip propagation time (BBR), which can maintain maximum transmission rate and minimum latency in a network link. However, the BBR algorithm was reported to cause serious round trip time (RTT) fairness problems by some evaluation experiments. The impact of the mismatch between pacing rate and bottleneck bandwidth caused by the asynchronous detection mechanism of BBR algorithm was analyzed to optimize the RTT fairness, and an optimized algorithm BBR-adaptive (BBR-A) was proposed based on pacing gain model. According to the relationship between RTT and pacing gain, a pacing gain adjustment model based on inverse proportional function was established, which replaces the fixed pacing gain coefficient in the original BBR algorithm. By interleaving the up and down pacing gain coefficients to adjust the pacing rate, each BBR flow could compete for bandwidth resources fairly. Experimental results of network simulator 3 (NS3) show that the channel utilization of BBR-A algorithm was slightly improved compared with BBR algorithm. In the experiment of RTT fairness, BBR-A reduced the throughput difference between different RTT flows, and Jain fairness index was at least 1.5 times higher than BBR algorithm with different buffer sizes and RTT differences. The retransmission rate of BBR-A algorithm was significantly reduced. By adaptively adjusting the pacing gain coefficient, the pacing rate between different flows was balanced, and the RTT fairness of BBR algorithm was improved. |
| Key words: congestion control BBR RTT fairness pacing gain adaptive algorithm |
