《计算机应用》唯一官方网站 ›› 2022, Vol. 42 ›› Issue (4): 1225-1234.DOI: 10.11772/j.issn.1001-9081.2021050722
• CCF第36届中国计算机应用大会 (CCF NCCA 2021) • 上一篇
收稿日期:
2021-04-22
修回日期:
2021-06-04
接受日期:
2021-06-08
发布日期:
2021-07-22
出版日期:
2022-04-10
通讯作者:
江勇
作者简介:
赖涵光(1995—),男,福建泉州人,硕士研究生,主要研究方向:端到端智能传输控制基金资助:
Hanguang LAI1, Qing LI2, Yong JIANG1()
Received:
2021-04-22
Revised:
2021-06-04
Accepted:
2021-06-08
Online:
2021-07-22
Published:
2022-04-10
Contact:
Yong JIANG
About author:
LAI Hanguang, born in 1995, M. S. candidate. His research interests include end-to-end intelligent transmission control.Supported by:
摘要:
针对轻量级基于学习的拥塞控制算法在某些场景下性能表现会出现断崖式下滑的问题,提出了一种基于场景变化的传输控制协议拥塞控制切换方案。首先,该方案模拟实时的网络环境;然后,根据实时的环境参数来识别场景;最后,将当前的拥塞控制算法切换至该场景下相对最优的轻量级基于学习的拥塞控制算法。实验结果表明,所提方案相较于原来使用单个拥塞控制算法的方案,例如测量瓶颈链路带宽和时延的拥塞控制(BBR)方案、面向性能的拥塞控制(PCC)方案等,可以使不同场景下的网络性能得到显著提升,总吞吐量增幅达到5%以上,总时延降幅达到10%以上。
中图分类号:
赖涵光, 李清, 江勇. 基于场景变化的传输控制协议拥塞控制切换方案[J]. 计算机应用, 2022, 42(4): 1225-1234.
Hanguang LAI, Qing LI, Yong JIANG. Transmission control protocol congestion control switching scheme based on scenario change[J]. Journal of Computer Applications, 2022, 42(4): 1225-1234.
拥塞控制算法 | 平均吞吐量/(Mb·s-1) | 平均时延/ms |
---|---|---|
BBR | 4.87 | 132 |
PCC-Vivace | 4.49 | 113 |
Copa | 4.78 | 106 |
Cubic | 4.63 | 141 |
表1 网络带宽为5 Mb/s和网络延迟为100 ms时的平均吞吐量与平均时延
Tab. 1 Average throughput and average delay with network bandwidth of 5 Mb/s and network delay of 100 ms
拥塞控制算法 | 平均吞吐量/(Mb·s-1) | 平均时延/ms |
---|---|---|
BBR | 4.87 | 132 |
PCC-Vivace | 4.49 | 113 |
Copa | 4.78 | 106 |
Cubic | 4.63 | 141 |
链路带宽/(Mb·s-1) | 网络延迟/ms | 随机丢包率/% | 使用算法 |
---|---|---|---|
1~100 | 1 | 0 | Vivace |
10 | 0 | Copa | |
100 | 0 | BBR | |
500 | 0 | BBR、Copa | |
5 | 1 | 0.1~1 | Vivace |
10 | 0.1~1 | Copa | |
100 | 0.1~1 | BBR | |
500 | 0.1~1 | BBR、Copa |
表2 考察吞吐量时的最优拥塞控制算法
Tab. 2 Optimal congestion control algorithm when focusing on throughput
链路带宽/(Mb·s-1) | 网络延迟/ms | 随机丢包率/% | 使用算法 |
---|---|---|---|
1~100 | 1 | 0 | Vivace |
10 | 0 | Copa | |
100 | 0 | BBR | |
500 | 0 | BBR、Copa | |
5 | 1 | 0.1~1 | Vivace |
10 | 0.1~1 | Copa | |
100 | 0.1~1 | BBR | |
500 | 0.1~1 | BBR、Copa |
链路带宽/(Mb·s-1) | 网络延迟/ms | 随机丢包率/% | 使用算法 |
---|---|---|---|
1~20 | 1 | 0 | Copa |
1 | 10 | 0 | Copa |
5 | 10 | 0 | Copa、Vivace |
20 | 10 | 0 | Vivace |
1 | 100 | 0 | Vivace |
100 | 0 | Copa | |
5~20 | 500 | 0 | Vivace |
500 | 0 | Copa | |
1 | 10 | 0.1 | Copa |
5 | 10 | 0.1 | Vivace |
20 | 10 | 0.1 | Vivace |
1 | 10 | 0.5~1 | Copa |
5 | 10 | 0.5~1 | Vivace |
20 | 10 | 0.5~1 | Copa |
表3 考察时延时的最优拥塞控制算法
Tab. 3 Optimal congestion control algorithm when focusing on delay
链路带宽/(Mb·s-1) | 网络延迟/ms | 随机丢包率/% | 使用算法 |
---|---|---|---|
1~20 | 1 | 0 | Copa |
1 | 10 | 0 | Copa |
5 | 10 | 0 | Copa、Vivace |
20 | 10 | 0 | Vivace |
1 | 100 | 0 | Vivace |
100 | 0 | Copa | |
5~20 | 500 | 0 | Vivace |
500 | 0 | Copa | |
1 | 10 | 0.1 | Copa |
5 | 10 | 0.1 | Vivace |
20 | 10 | 0.1 | Vivace |
1 | 10 | 0.5~1 | Copa |
5 | 10 | 0.5~1 | Vivace |
20 | 10 | 0.5~1 | Copa |
链路带宽/(Mb·s-1) | 网络延迟/ms | 随机丢包率/% | 使用算法 |
---|---|---|---|
1~100 | 100 | 0 | Vivace |
100 | 0.1~1 | Copa |
表4 考察公平性时的最优拥塞控制算法
Tab. 4 Optimal congestion control algorithm when focusing on fairness
链路带宽/(Mb·s-1) | 网络延迟/ms | 随机丢包率/% | 使用算法 |
---|---|---|---|
1~100 | 100 | 0 | Vivace |
100 | 0.1~1 | Copa |
链路带宽/(Mb·s-1) | 网络延迟/ms | 随机丢包率/% | 使用算法 |
---|---|---|---|
1 | 1 | 0~1 | Copa |
5 | 1 | 0~0.5 | Vivace |
5 | 1 | 1 | BBR |
20 | 1 | 0~1 | BBR |
100 | 1 | 0~0.1 | BBR |
1 | 0.5~1 | Vivace |
表5 考察TCP友好性时的最优拥塞控制算法
Tab. 5 Optimal congestion control algorithm when focusing onTCP friendliness
链路带宽/(Mb·s-1) | 网络延迟/ms | 随机丢包率/% | 使用算法 |
---|---|---|---|
1 | 1 | 0~1 | Copa |
5 | 1 | 0~0.5 | Vivace |
5 | 1 | 1 | BBR |
20 | 1 | 0~1 | BBR |
100 | 1 | 0~0.1 | BBR |
1 | 0.5~1 | Vivace |
组别 | 时刻 | 变化时刻/s | 带宽/(Mb·s-1) | 延迟/ms |
---|---|---|---|---|
1 | t1 | 8 | 193 | 5 |
t2 | 243 | 45 | 17 | |
t3 | 364 | 7 | 91 | |
t4 | 445 | 5 | 40 | |
2 | t1 | 48 | 19 | 22 |
t2 | 178 | 2 | 16 | |
t3 | 285 | 275 | 1 | |
t4 | 438 | 2 | 7 | |
3 | t1 | 50 | 1 | 18 |
t2 | 139 | 172 | 47 | |
t3 | 368 | 2 | 13 | |
t4 | 447 | 82 | 2 |
表6 三组实验随机产生的参数
Tab. 6 Randomly generated parameters in 3 sets of experiments
组别 | 时刻 | 变化时刻/s | 带宽/(Mb·s-1) | 延迟/ms |
---|---|---|---|---|
1 | t1 | 8 | 193 | 5 |
t2 | 243 | 45 | 17 | |
t3 | 364 | 7 | 91 | |
t4 | 445 | 5 | 40 | |
2 | t1 | 48 | 19 | 22 |
t2 | 178 | 2 | 16 | |
t3 | 285 | 275 | 1 | |
t4 | 438 | 2 | 7 | |
3 | t1 | 50 | 1 | 18 |
t2 | 139 | 172 | 47 | |
t3 | 368 | 2 | 13 | |
t4 | 447 | 82 | 2 |
时刻 | 第1组 | 第2组 | 第3组 | |||
---|---|---|---|---|---|---|
吞吐量 | 时延 | 吞吐量 | 时延 | 吞吐量 | 时延 | |
t1 | Copa | PCC-Vivace | BBR | PCC-Vivace | Copa | Copa |
t2 | BBR | Copa | BBR | Copa | BBR | Copa |
t3 | BBR | Copa | PCC-Vivace | Copa | Copa | Copa |
t4 | BBR | Copa | Copa | Copa | PCC-Vivace | Copa |
表7 四个随机时刻下3组实验根据环境参数求得的拥塞控制算法
Tab. 7 Obtained congestion control algorithm based on environmental parameters at 4 random moments in 3 sets of experiments
时刻 | 第1组 | 第2组 | 第3组 | |||
---|---|---|---|---|---|---|
吞吐量 | 时延 | 吞吐量 | 时延 | 吞吐量 | 时延 | |
t1 | Copa | PCC-Vivace | BBR | PCC-Vivace | Copa | Copa |
t2 | BBR | Copa | BBR | Copa | BBR | Copa |
t3 | BBR | Copa | PCC-Vivace | Copa | Copa | Copa |
t4 | BBR | Copa | Copa | Copa | PCC-Vivace | Copa |
1 | ALLMAN M, PAXSON V, STEVENS W. TCP congestion control: RFC 2581 [S]. Reston, VA: Internet Society, 1999-04. |
2 | SIVARAMAN A, WINSTEIN K, THAKER P, et al. An experimental study of the learnability of congestion control[C]// Proceedings of 2014 ACM Conference on SIGCOMM. New York: ACM, 2014:479-490. 10.1145/2740070.2626324 |
3 | JAY N, ROTMAN N H, GODFREY P B, et al. A deep reinforcement learning perspective on internet congestion control[C]// Proceedings of the 36th International Conference on Machine Learning. New York: JMLR.org, 2019:3050-3059. |
4 | ABBASLOO S, YEN C Y, CHAO H J. Classic meets modern: a pragmatic learning-based congestion control for the internet[C]// Proceedings of the 2020 Annual Conference of the ACM Special Interest Group on Data Communication on the Applications, Technologies, Architectures, and Protocols for Computer Communication. New York: ACM, 2020:632-647. 10.1145/3387514.3405892 |
5 | CARDWELL N, CHENG Y, GUNN C S, et al. BBR: congestion-based congestion control[J]. Communications of the ACM, 2017, 60(2): 58-66. 10.1145/3009824 |
6 | DONG M, LI Q X, ZARCHY D, et al. PCC: re-architecting congestion control for consistent high performance[C]// Proceedings of the 12th USENIX Symposium on Networked Systems Design and Implementation. Berkeley: USENIX Association, 2015:395-408. |
7 | ARUN V, BALAKRISHNAN H. Copa: congestion control combining objective optimization with window adjustments[C]// Proceedings of the 15th USENIX Symposium on Networked Systems Design and Implementation. Berkeley: USENIX Association, 2018:329-342. 10.1145/3232755.3232783 |
8 | ALIZADEH M, GREENBERG A, MALTZ D A, et al. Data center TCP (DCTCP)[C]// Proceedings of the 2010 ACM SIGCOMM Conference. New York: ACM, 2010:63-74. 10.1145/1851182.1851192 |
9 | SIVARAMAN A, WINSTEIN K, SUBRAMANIAN S, et al. No silver bullet: extending SDN to the data plane[C]// Proceedings of 12th ACM Workshop on Hot Topics in Networks. New York: ACM, 2013:No.19. 10.1145/2535771.2535796 |
10 | DONG M, MENG T, ZARCHY D, et al. PCC Vivace: online-learning congestion control[C]// Proceedings of the 15th USENIX Symposium on Networked Systems Design and Implementation. Berkeley: USENIX Association, 2018:343-356. |
11 | SCHAPIRA M, WINSTEIN K. Congestion-control throwdown[C]// Proceedings of the 16th ACM Workshop on Hot Topics in Networks. New York: ACM, 2017:122-128. 10.1145/3152434.3152446 |
12 | GETTYS J, NICHOLS K. Bufferbloat: dark buffers in the Internet[J]. Communications of the ACM, 2012, 55( 1):57-65. 10.1145/2063176.2063196 |
13 | HA S, RHEE I, XU L S. CUBIC: a new TCP-friendly high-speed TCP variant[J]. ACM SIGOPS Operating Systems Review, 2008, 42(5):64-74. 10.1145/1400097.1400105 |
14 | SCHWARZ M, SAUER C, DADUNA H, et al. M/M/1 Queueing systems with inventory[J]. Queueing Systems, 2006, 54(1): 55-78. 10.1007/s11134-006-8710-5 |
15 | WINSTEIN K, BALAKRISHNAN H. TCP ex Machina: computer-generated congestion control[C]// Proceedings of the 2013 ACM SIGCOMM Conference. New York: ACM, 2013:123-134. 10.1145/2486001.2486020 |
16 | FRANÇOIS-LAVET V, HENDERSON P, ISLAM R, et al. An introduction to deep reinforcement learning[J]. Foundations and Trends in Machine Learning, 2018, 11(3/4):219-354. 10.1561/2200000071 |
17 | OSBAND I, BLUNDELL C, PRITZEL A, et al. Deep exploration via bootstrapped DQN[C]// Proceedings of the 30th International Conference on Neural Information Processing Systems. Red Hook, NY: Curran Associates Inc., 2016:4033-4041. |
18 | LILLICRAP T P, HUNT J J, PRITZEL A, et al. Continuous control with deep reinforcement learning[EB/OL]. (2019-07-05) [2021-03-03]. . |
19 | LI Z B, LIU P, XU C C, et al. Reinforcement learning-based variable speed limit control strategy to reduce traffic congestion at freeway recurrent bottlenecks[J]. IEEE Transactions on Intelligent Transportation Systems, 2017, 18(11):3204-3217. 10.1109/tits.2017.2687620 |
20 | KONDA V R, TSITSIKLIS J N. Actor-critic algorithms[C]// Proceedings of the 1999 Annual Conference on Neural Information Processing Systems. Cambridge: MIT Press, 1999:1008-1014. |
21 | YAN F Y, MA J, HILL G D, et al. Pantheon: the training ground for Internet congestion-control research[C]// Proceedings of the 2018 USENIX Technical Conference. Berkeley: USENIX Association, 2018:731-743. |
22 | GIESSLER A, HÄNLE J, KÖNIG A, et al. Free buffer allocation — an investigation by simulation[J]. Computer Networks, 1978, 2(3):191-208. 10.1016/0376-5075(78)90028-4 |
23 | GAIL R, KLEINROCK L. An invariant property of computer network power[C]// Proceedings of the 1981 International Conference on Communications. Piscataway: IEEE, 1981:63.1.1-63.1.5. |
24 | SCHAUL T, QUAN J, ANTONOGLOU I, et al. Prioritized experience replay[EB/OL]. (2016-02-25) [2021-03-03].. |
25 | 沈耿彪,李清,江勇,等. 数据中心网络负载均衡问题研究[J]. 软件学报, 2020, 31(7):2221-2244. 10.13328/j.cnki.jos.006050 |
SHEN G B, LI Q, JIANG Y, et al. Research on load balancing in data center networks[J]. Journal of Software, 2020, 31(7):2221-2244. 10.13328/j.cnki.jos.006050 |
[1] | 史杨潇, 章军, 陈鹏, 王兵. 基于轻量级网络的钢铁表面缺陷分类[J]. 计算机应用, 2021, 41(6): 1836-1841. |
[2] | 吴恺凡, 殷新春. 基于随机运算符的轻量级匿名射频识别系统双向认证协议[J]. 计算机应用, 2021, 41(6): 1621-1630. |
[3] | 胡嵽, 冯子亮. 基于深度学习的轻量级道路图像语义分割算法[J]. 计算机应用, 2021, 41(5): 1326-1331. |
[4] | 祝永晋, 尹飞, 豆龙龙, 吴昆, 张志伟, 钱柱中. 基于前向纠错的自适应网络传输机制[J]. 计算机应用, 2021, 41(3): 825-832. |
[5] | 顾桐, 许国良, 李万林, 李家浩, 王志愿, 雒江涛. 基于集成LightGBM和贝叶斯优化策略的房价智能评估模型[J]. 计算机应用, 2020, 40(9): 2762-2767. |
[6] | 朱怡, 宁振虎, 周艺华. 基于视觉特征的仿冒域名轻量级检测技术[J]. 计算机应用, 2020, 40(8): 2279-2285. |
[7] | 强保华, 翟艺杰, 陈金龙, 谢武, 郑虹, 王学文, 张世豪. 基于改进CPMs和SqueezeNet的轻量级人体骨骼关键点检测模型[J]. 计算机应用, 2020, 40(6): 1806-1811. |
[8] | 赵羽龙, 牛保宁, 李鹏, 樊星. 区块链增强型轻量级节点模型[J]. 计算机应用, 2020, 40(4): 942-946. |
[9] | 王继霄, 李阳, 王家宝, 苗壮, 张洋硕. 基于SqueezeNet的轻量级图像融合方法[J]. 计算机应用, 2020, 40(3): 837-841. |
[10] | 龚锐, 丁胜, 章超华, 苏浩. 基于深度学习的轻量级和多姿态人脸识别方法[J]. 计算机应用, 2020, 40(3): 704-709. |
[11] | 皮家甜, 杨杰之, 杨琳希, 彭明杰, 邓雄, 赵立军, 唐万梅, 吴至友. 基于多模态特征融合的轻量级人脸活体检测方法[J]. 计算机应用, 2020, 40(12): 3658-3665. |
[12] | 史梦安, 陆振宇. 基于YOLO-tiny-RFB模型的电站旋钮开关状态识别[J]. 计算机应用, 2020, 40(12): 3679-3686. |
[13] | 陈放, 刘晓瑞, 杨明业. 基于活体检测和身份认证的人脸识别安防系统[J]. 计算机应用, 2020, 40(12): 3666-3672. |
[14] | 陈力, 丁世飞, 于文家. 基于跨通道交叉融合和跨模块连接的轻量级卷积神经网络[J]. 计算机应用, 2020, 40(12): 3451-3457. |
[15] | 刘宗甫, 袁征, 赵晨曦, 朱亮. 对PICO算法基于可分性的积分攻击[J]. 计算机应用, 2020, 40(10): 2967-2972. |
阅读次数 | ||||||
全文 |
|
|||||
摘要 |
|
|||||