A new TCP congestion control supporting RTT-fairness

Kazumine Ogura, Yohei Nemoto, Zhou Su, Jiro Katto

    Research output: Contribution to journalArticle

    6 Citations (Scopus)

    Abstract

    This paper focuses on RTT-fairness of multiple TCP flows over the Internet, and proposes a new TCP congestion control named "HRF (Hybrid RTT-Fair)-TCP". Today, it is a serious problem that the flows having smaller RTT utilize more bandwidth than others when multiple flows having different RTT values compete in the same network. This means that a user with longer RTT may not be able to obtain sufficient bandwidth by the current methods. This RTT fairness issue has been discussed in many TCP papers. An example is CR (Constant Rate) algorithm, which achieves RTT-fairness by multiplying the square of RTT value in its window increment phase against TCP-Reno. However, the method halves its windows size same as TCP-Reno when a packet loss is detected. This makes worse its efficiency in certain network cases. On the other hand, recent proposed TCP versions essentially require throughput efficiency and TCPfriendliness with TCP-Reno. Therefore, we try to keep these advantages in our TCP design in addition to RTT-fairness. In this paper, we make intuitive analytical models in which we separate resource utilization processes into two cases: utilization of bottleneck link capacity and that of buffer space at the bottleneck link router. These models take into account three characteristic algorithms (Reno, Constant Rate, Constant Increase) in window increment phase where a sender receives an acknowledgement successfully. Their validity is proved by both simulations and implementations. From these analyses, we propose HRF-TCP which switches two modes according to observed RTT values and achieves RTT fairness. Experiments are carried out to validate the proposed method. Finally, HRF-TCP outperforms conventional methods in RTT-fairness, efficiency and friendliness with TCP-Reno.

    Original languageEnglish
    Pages (from-to)523-531
    Number of pages9
    JournalIEICE Transactions on Information and Systems
    VolumeE95-D
    Issue number2
    DOIs
    Publication statusPublished - 2012 Feb

    Fingerprint

    Bandwidth
    Congestion control (communication)
    Packet loss
    Routers
    Analytical models
    Rate constants
    Switches
    Throughput
    Internet
    Experiments

    Keywords

    • RTT-fairness
    • Transport protocol

    ASJC Scopus subject areas

    • Electrical and Electronic Engineering
    • Software
    • Artificial Intelligence
    • Hardware and Architecture
    • Computer Vision and Pattern Recognition

    Cite this

    A new TCP congestion control supporting RTT-fairness. / Ogura, Kazumine; Nemoto, Yohei; Su, Zhou; Katto, Jiro.

    In: IEICE Transactions on Information and Systems, Vol. E95-D, No. 2, 02.2012, p. 523-531.

    Research output: Contribution to journalArticle

    Ogura, Kazumine ; Nemoto, Yohei ; Su, Zhou ; Katto, Jiro. / A new TCP congestion control supporting RTT-fairness. In: IEICE Transactions on Information and Systems. 2012 ; Vol. E95-D, No. 2. pp. 523-531.
    @article{61b0d8b7205c4d4aa8ab0f94fceef3c5,
    title = "A new TCP congestion control supporting RTT-fairness",
    abstract = "This paper focuses on RTT-fairness of multiple TCP flows over the Internet, and proposes a new TCP congestion control named {"}HRF (Hybrid RTT-Fair)-TCP{"}. Today, it is a serious problem that the flows having smaller RTT utilize more bandwidth than others when multiple flows having different RTT values compete in the same network. This means that a user with longer RTT may not be able to obtain sufficient bandwidth by the current methods. This RTT fairness issue has been discussed in many TCP papers. An example is CR (Constant Rate) algorithm, which achieves RTT-fairness by multiplying the square of RTT value in its window increment phase against TCP-Reno. However, the method halves its windows size same as TCP-Reno when a packet loss is detected. This makes worse its efficiency in certain network cases. On the other hand, recent proposed TCP versions essentially require throughput efficiency and TCPfriendliness with TCP-Reno. Therefore, we try to keep these advantages in our TCP design in addition to RTT-fairness. In this paper, we make intuitive analytical models in which we separate resource utilization processes into two cases: utilization of bottleneck link capacity and that of buffer space at the bottleneck link router. These models take into account three characteristic algorithms (Reno, Constant Rate, Constant Increase) in window increment phase where a sender receives an acknowledgement successfully. Their validity is proved by both simulations and implementations. From these analyses, we propose HRF-TCP which switches two modes according to observed RTT values and achieves RTT fairness. Experiments are carried out to validate the proposed method. Finally, HRF-TCP outperforms conventional methods in RTT-fairness, efficiency and friendliness with TCP-Reno.",
    keywords = "RTT-fairness, Transport protocol",
    author = "Kazumine Ogura and Yohei Nemoto and Zhou Su and Jiro Katto",
    year = "2012",
    month = "2",
    doi = "10.1587/transinf.E95.D.523",
    language = "English",
    volume = "E95-D",
    pages = "523--531",
    journal = "IEICE Transactions on Information and Systems",
    issn = "0916-8532",
    publisher = "Maruzen Co., Ltd/Maruzen Kabushikikaisha",
    number = "2",

    }

    TY - JOUR

    T1 - A new TCP congestion control supporting RTT-fairness

    AU - Ogura, Kazumine

    AU - Nemoto, Yohei

    AU - Su, Zhou

    AU - Katto, Jiro

    PY - 2012/2

    Y1 - 2012/2

    N2 - This paper focuses on RTT-fairness of multiple TCP flows over the Internet, and proposes a new TCP congestion control named "HRF (Hybrid RTT-Fair)-TCP". Today, it is a serious problem that the flows having smaller RTT utilize more bandwidth than others when multiple flows having different RTT values compete in the same network. This means that a user with longer RTT may not be able to obtain sufficient bandwidth by the current methods. This RTT fairness issue has been discussed in many TCP papers. An example is CR (Constant Rate) algorithm, which achieves RTT-fairness by multiplying the square of RTT value in its window increment phase against TCP-Reno. However, the method halves its windows size same as TCP-Reno when a packet loss is detected. This makes worse its efficiency in certain network cases. On the other hand, recent proposed TCP versions essentially require throughput efficiency and TCPfriendliness with TCP-Reno. Therefore, we try to keep these advantages in our TCP design in addition to RTT-fairness. In this paper, we make intuitive analytical models in which we separate resource utilization processes into two cases: utilization of bottleneck link capacity and that of buffer space at the bottleneck link router. These models take into account three characteristic algorithms (Reno, Constant Rate, Constant Increase) in window increment phase where a sender receives an acknowledgement successfully. Their validity is proved by both simulations and implementations. From these analyses, we propose HRF-TCP which switches two modes according to observed RTT values and achieves RTT fairness. Experiments are carried out to validate the proposed method. Finally, HRF-TCP outperforms conventional methods in RTT-fairness, efficiency and friendliness with TCP-Reno.

    AB - This paper focuses on RTT-fairness of multiple TCP flows over the Internet, and proposes a new TCP congestion control named "HRF (Hybrid RTT-Fair)-TCP". Today, it is a serious problem that the flows having smaller RTT utilize more bandwidth than others when multiple flows having different RTT values compete in the same network. This means that a user with longer RTT may not be able to obtain sufficient bandwidth by the current methods. This RTT fairness issue has been discussed in many TCP papers. An example is CR (Constant Rate) algorithm, which achieves RTT-fairness by multiplying the square of RTT value in its window increment phase against TCP-Reno. However, the method halves its windows size same as TCP-Reno when a packet loss is detected. This makes worse its efficiency in certain network cases. On the other hand, recent proposed TCP versions essentially require throughput efficiency and TCPfriendliness with TCP-Reno. Therefore, we try to keep these advantages in our TCP design in addition to RTT-fairness. In this paper, we make intuitive analytical models in which we separate resource utilization processes into two cases: utilization of bottleneck link capacity and that of buffer space at the bottleneck link router. These models take into account three characteristic algorithms (Reno, Constant Rate, Constant Increase) in window increment phase where a sender receives an acknowledgement successfully. Their validity is proved by both simulations and implementations. From these analyses, we propose HRF-TCP which switches two modes according to observed RTT values and achieves RTT fairness. Experiments are carried out to validate the proposed method. Finally, HRF-TCP outperforms conventional methods in RTT-fairness, efficiency and friendliness with TCP-Reno.

    KW - RTT-fairness

    KW - Transport protocol

    UR - http://www.scopus.com/inward/record.url?scp=84856407046&partnerID=8YFLogxK

    UR - http://www.scopus.com/inward/citedby.url?scp=84856407046&partnerID=8YFLogxK

    U2 - 10.1587/transinf.E95.D.523

    DO - 10.1587/transinf.E95.D.523

    M3 - Article

    AN - SCOPUS:84856407046

    VL - E95-D

    SP - 523

    EP - 531

    JO - IEICE Transactions on Information and Systems

    JF - IEICE Transactions on Information and Systems

    SN - 0916-8532

    IS - 2

    ER -