Minimum TCP throughput guarantee on minimum rate guaranteed networks

  • Authors:
  • Jeongsik In;Kyeong Hur;Jinwoo Park;Kyun Hyon Tchah;Dooseop Eom

  • Affiliations:
  • School of Electrical Engineering, Korea University, 1, 5-ka, Anam-dong Sungbuk-ku, Seoul 136-701, South Korea;School of Electrical Engineering, Korea University, 1, 5-ka, Anam-dong Sungbuk-ku, Seoul 136-701, South Korea;School of Electrical Engineering, Korea University, 1, 5-ka, Anam-dong Sungbuk-ku, Seoul 136-701, South Korea;School of Electrical Engineering, Korea University, 1, 5-ka, Anam-dong Sungbuk-ku, Seoul 136-701, South Korea;School of Electrical Engineering, Korea University, 1, 5-ka, Anam-dong Sungbuk-ku, Seoul 136-701, South Korea

  • Venue:
  • Computer Communications
  • Year:
  • 2004

Quantified Score

Hi-index 0.24

Visualization

Abstract

The performance of Transmission Control Protocol (TCP) entirely depends on the network traffic conditions. It will be greatly advantageous if the throughput of the TCP is always maintained above a certain level. However, it is difficult to guarantee minimum throughput for each TCP connection even when enough bandwidth is reserved for it. This is because the TCP congestion control algorithm makes its transmission rate fluctuate so that, in some cases, the connection cannot fully utilize the portion of the bandwidth reserved for it. We propose a slightly modified traffic shaper that can help maintain the TCP transmission rate. It normally regulates traffic just as the conventional shapers, but when necessary, it adjusts its transmission rate depending on its own state, namely its queue length and packet-receiving rate, without explicit notification of the network conditions. It only requires a simple modification of the conventional shapers. However, this simple modification can improve TCP performance significantly and help guarantee minimum throughput when the network guarantees minimum bandwidth for the connection. Simulation results are presented for performance evaluation. The proposed scheme is practical in that it does not require end-host modifications. Also, it has scalability since it requires per-flow management only in the network access routers and not in the core routers.