SCIENTIA SINICA Informationis, Volume 49, Issue 10: 1321-1332(2019) https://doi.org/10.1360/N112019-00010

Active queue management algorithm for time delay demand

More info
  • ReceivedJan 16, 2019
  • AcceptedJun 18, 2019
  • PublishedOct 17, 2019


With the popularization and development of network intelligent terminals, additional requirements and challenges have arisen concerning network congestion control and the quality of service based on delay and jitter. Existing network intermediate devices primarily rely on active queue management to control the delay and solve congestion. This paper proposes an active queue management algorithm focusing on time delay demands to solve the large number of packet loss problems in existing active queue management algorithms. Packet loss problems are caused by long delays and unequal network resource allocation. In the TD-AQM algorithm, the network measurement feedback is obtained from the end system and intermediate equipment is returned to the data packet as the basis for queue management. The delay requirement is written in the option field of the IP layer of the data packet, i.e., the longest time of the message in the router. To avoid the flood peak effect, TD-AQM constrains the down lookup range and limits the occupancy accuracy of the lattice. The queue location of packets is determined according to the delay requirements. The experimental results show that TD-AQM can effectively maintain the stability of the queue, ensure that data packets are queued according to the delay requirements, and guarantee an overall low latency compared to other algorithms.

Funded by






[1] Grazia C A, Patriciello N, Klapez M, et al. Mitigating congestion and Bufferbloat on satellite networks through a rate-based AQM. In: Proceedings of the IEEE International Conference on Communications, Paris, 2017. 1--6. Google Scholar

[2] Casoni M, Grazia C A, Klapez M. How to avoid TCP congestion without dropping packets: An effective AQM called PINK. Comput Commun, 2017, 103: 49-60 CrossRef Google Scholar

[3] Sanjeev Patel. Performance analysis on RED for stabilized queue. In: Proceedings of the IEEE Seventh International Conference on Contemporary Computing, Noida, 2014. 306--311. Google Scholar

[4] Sanjeev Patel. Performance analysis and modeling of congestion control algorithms based on active queue management. In: Proceedings of the IEEE International Conference on Signal Processing and Communication, Vancouver, 2013. 449--454. Google Scholar

[5] Tarun Jain, Annappa B, Mohit P, et al. Performance evaluation of CoDel for active queue management in wired-cum-wireless networks. In: Proceedings of the IEEE Fourth International Conference on Advanced Computing Communication Technologies, Rohtak, 2014. 381--385. Google Scholar

[6] Tahiliani M P, Shet K C, Basavaraju T G. CARED: Cautious Adaptive RED gateways for TCP/IP networks. J Network Comput Appl, 2012, 35: 857-864 CrossRef Google Scholar

[7] Liu W Y, Liu B, Zou X L. Congestion control algorithm based on dynamic threshold. Application Research of Computers, 2013, 30:3459-3461. Google Scholar

[8] Alsaaidah A, Zalisham M, Fadzli M, et al. Gentle-BLUE: A New Method for Active Queue Management. In: Proceedings of the International Conference on Advanced Computer Science Applications and Technologies, Amman, 2015. 67--72. Google Scholar

[9] Nichols K, Jacobson V. Controlling queue delay. Commun ACM, 2012, 55: 42 CrossRef Google Scholar

[10] Tian Z H, Yu X Z, Zhang H L, et al. A real-time network intrusion forensics method based on evidence reasoning network. Chin J Comput, 2014, (05). Google Scholar

[11] Zhu J, Luo T, Yang L. An average queue-length-difference-based congestion detection algorithm in TCP/AQM network. Int J Adapt Control Signal Process, 2018, 32: 742-752 CrossRef Google Scholar

[12] Bisoy S K, Pattnaik P K, Pati B. Design and analysis of a stable AQM controller for network congestion control. IJCNDS, 2018, 20: 143 CrossRef Google Scholar

[13] Iskandar M N. Active Queue Management (AQM) Performance Analysis Based On Controlled Delay (CoDel) Against Bufferbloat On Real-Time Application: Procedia Computer Science. 2017, 2:119. Google Scholar

[14] Sheikhan M, Shahnazi R, Hemmati E. Adaptive active queue management controller for TCP communication networks using PSO-RBF models. Neural Comput Applic, 2013, 22: 933-945 CrossRef Google Scholar

[15] Okokpujie K, Chukwu E C, Noma-Osaghae E. Novel Active Queue Management Scheme for Routers in Wireless Networks. IRECAP, 2018, 8: 52 CrossRef Google Scholar

[16] Hamidian H, Beheshti M T H. A robust fractional-order PID controller design based on active queue management for TCP network. Int J Syst Sci, 2018, 49: 211-216 CrossRef Google Scholar

[17] Tahiliani M P, Shet K C. Analysis of cautious adaptive RED (CARED). In: Proceedings of the International Conference on Advances in Computing, Communications and Informatics, Mysore, 2013. 1029--1034. Google Scholar

[18] Mühlenthaler M, Wanka R. Fairness in academic course timetabling. Ann Oper Res, 2016, 239: 171-188 CrossRef Google Scholar

  • Figure 1

    Frame of TD-AQM

  • Figure 2

    Queue structure

  • Figure 3

    Example of flood peak effect

  • Figure 4

    Queue management policy of TD-AQM

  • Figure 5

    Experimental topology

  • Figure 6

    (Color online) Performance comparison.(a) Throughput rate; (b) fairness

  • Figure 7

    (Color online) Queue stability comparison.(a) TD-AQM; (b) other algorithms

  • Table 1   Comparison of time demand satisfaction
    Number of flows Standard normal distribution Uniform distribution No time delay demand
    20 0.81 0.82 0.91
    40 0.84 0.81 0.85
    60 0.90 0.78 0.82
    100 0.86 0.77 0.81

    Algorithm 1 Constrain down lookup range

    Require:Time delay demand TimeDemand, Average delay of single packet $t$;

    Output:Extreme searchable downward position last_ position;

    ${last\underline{ }position}\Leftarrow1$;

    if ${\rm~TimeDemand}~=~0$ then

    ${pre\underline{ }position}\Leftarrow1$;

    ${last\underline{ }position}\Leftarrow1$;


    ${pre\underline{ }position}=\frac{{\rm~TimeDemand}}{t}$;

    end if

    if ${pre\underline{ }position}\leq100$ then

    ${last\underline{ }position}\Leftarrow1$;

    end if

    if ${pre\underline{ }position}>100$ then

    ${last\underline{ }position}\Leftarrow50+50\times\frac{1}{1+({pre\underline{ }position}-100)}$;

    end if


    Algorithm 2 Limit virtual occupancy accuracy

    Require:Estimate enqueue position pre_ position, extreme searchable downward position last_ position, lattice $n$, instantaneous queue length $q_i$, average queue length $q_{\rm~avg}$;

    Input:Lattice $n$ hit rate $P$;


    if $n>{pre\underline{ }position}~\&\&~n<{last\underline{ }position}$ then


    end if

    if ${pre\underline{ }position}>100$ then

    $P=\frac{1}{{pre\underline{ }position}-n}\times\log_{q_i}q_{\rm~avg}$;

    end if

Copyright 2020 Science China Press Co., Ltd. 《中国科学》杂志社有限责任公司 版权所有

京ICP备18024590号-1       京公网安备11010102003388号