Elsevier

Computer Communications

Volume 31, Issue 10, 25 June 2008, Pages 2242-2256
Computer Communications

TCP friendly protocols for media streams over heterogeneous wired–wireless networks

https://doi.org/10.1016/j.comcom.2008.02.012Get rights and content

Abstract

The growing need for Internet friendly streaming protocols prompted us to develop IFTP (Internet Friendly Transport Protocol). IFTP is a protocol with an inherent rate-based flow control mechanism designed to emulate TCP’s window-based flow control mechanism. In this paper, we present two proposals to improve the performance of IFTP in networks with wireless links. The first is IFTP-W, which specifies an error-sensitive section. Bit errors not occurring in this section are ignored. The second is IFTPV, which distinguishes between losses due to congestion and those due to the wireless link. In this paper, we present the performance of IFTP-W with varying error-sensitive section lengths on networks with varying link qualities. We also analyze the performance of IFTPV in extensive scenarios with various network conditions. The results show that both proposals maintain the TCP-friendliness and fairness properties, but perform better than IFTP on wireless networks.

Introduction

The growth in the popularity of multimedia applications and real-time games is leading to an increased usage of continuous media protocols which lack end-to-end congestion control mechanisms. The lack of flow control inherent to these protocols can lead to a monopolization of the available bandwidth as these protocols are not concerned with reliable data delivery. This problem, called the Internet friendly problem or the TCP-friendly problem has not fully manifested itself, but it poses as a potentially substantial problem that should be addressed before requiring immediate attention. Not only could this problem cause inequitable bandwidth distribution, but it also threatens congestion collapse [1].

There are several solutions to counteracting the effect of flows unresponsive to path congestion, including active queue management and flow-based packet scheduling at the router level. However, if each flow is aware of network conditions, such monitoring and management of each flow at the router level is not required. Several methods of informing flows of imminent packet loss such as Explicit Congestion Notification (ECN) have been developed, but they are only effective for flows that already contain some form of congestion control. The Internet Friendly Transport-level Protocol (IFTP) [2], is one solution to the Internet friendly problem. IFTP combines the streaming qualities of UDP and the congestion control properties of TCP to create a continuous media protocol capable of responding appropriately to congestion.

As the usage of wireless networks increases, the necessity of developing a streaming protocol capable of functioning over a wireless medium also increases. Wireless networks have the eminent problems of high bit error rate (BER) and small bandwidth. These issues adversely affect the performance of protocols that do not account for them. For example, as most versions of TCP interpret packet drops as a signal of congestion, a packet drop caused by bit corruption is mistakenly identified as an indication of network congestion. On networks with high BER, this can easily cripple the connection.

In this paper, we propose transport protocols for media streaming that is aware of the TCP-friendly problem and the limitations of wireless networks. We modified the TCP-friendly protocol (IFTP) that we previously developed [2] to enhance its performance on wireless networks. Like TCP, IFTP interprets packet loss as network congestion and reacts to it by decreasing its sending rate. This negatively and unnecessarily affects the performance of IFTP as packet loss in wireless networks is mainly related to high BER rather than congestion. To improve the performance of IFTP, we first propose IFTP-Wireless or IFTP-W – a protocol that distinguishes between error-sensitive and error-insensitive data. If an IFTP-W packet is corrupted due to the high BER of the wireless medium, it is dropped only if the error occurred in error-sensitive data. IFTP-W adjusts its sending rate according to IFTP’s congestion avoidance protocol if error-sensitive data is corrupted. Our simulation results show that IFTP-W performs better than IFTP on wireless networks and maintains IFTP’s TCP-friendliness property. We also study the effect of error-sensitive length of IFTP-W. Second, we present another transport protocol, IFTPV, to solve the TCP-friendly problem on wireless networks. This protocol attempts to differentiate between errors due to congestions and those due to the wireless link. We study the performance of IFTPV through extensive simulations in various network scenarios.

The rest of the paper is organized as follows: Section 2 presents the background for the paper, while Section 3 presents previous research in this area. In Section 4, we briefly explain IFTP protocol. Section 5 explains our discrete event network simulator, network topology and performance metrics. Section 6 shows the performance of IFTP on wireless networks which illustrates the need for protocols that are designed with wireless problems in mind. Section 7, introduces the IFTP-W protocol, while Section 8 demonstrates its performance on wireless networks through simulation using several performance metrics. Section 9 studies the effect of varying the length of IFTP-W’s error-sensitive part. Section 10, introduces IFTPV and its performance evaluation. Finally Section 11 concludes the paper and provides suggestions for future work.

Section snippets

Background

In this section we give a brief overview of UDP and TCP. UDP is a connectionless unreliable protocol designed for multimedia and other streaming applications. UDP transmits packets at a set rate. The rate is realized by calculating an appropriate interpacket gap. The interpacket gap is generally simulated via an exponential distribution to model the actual performance of UDP. Thus, our UDP server transmits packets at an average rate with an exponentially distributed interpacket gap and does not

Previous research

Past research has included suggestions for both TCP-friendly congestion control mechanisms and improved performance over wireless networks as independent topics. A good survey of TCP-friendly protocols can be found in [6]. One example protocol, TCP Friendly Rate Control (TFRC) [7] attempts to approximate the bandwidth usage of TCP, though it strives for a smoother transmission rate, and is thus less responsive to changes in network conditions than IFTP.

Several suggestions have been proposed to

IFTP

In this section, we briefly explain the IFTP protocol as it is the base of our proposal protocols. More details can be found in [2]. IFTP is an Internet protocol that was designed to provide an Internet friendly transmission protocol for multimedia applications. By Internet friendly we mean that it scales its transmission rate in a fashion similar to TCP as network congestion is introduced. Such a protocol was needed because most multimedia applications are transmitted using UDP, which is

Simulator, topology and metrics

In the next section, we present the performance of IFTP over wireless and follow it with IFTP-W and IFTPV. Since all the performance measures reported in this paper are done through simulations, we first present our network simulator, the network topology and the performance metrics used in the rest of the paper.

IFTP over wireless networks

We studied the performance of IFTP on wireless networks through simulation. Fig. 2, Fig. 3, Fig. 4, Fig. 5 show our simulation results for the four mentioned performance measures.

As expected, UDP reacts less severely to the change of BER than either IFTP or TCP as shown in Fig. 2. This follows expectations because UDP is devoid of flow control, and thus any throughput reduction when encountering dropped packets is caused solely by the loss of the segments from the network. The sending rate is

IFTP-W

IFTP-W [14], [15] is an end-to-end congestion control protocol we designed to solve IFTP’s problems on wireless networks illustrated in Section 6. IFTP-W is a TCP-friendly protocol for media streams that allows for applications to choose a section of a packet to be verified by a checksum, with the remaining portion deemed error-insensitive and not checked for corrupted bits. One corrupted bit in a video or audio stream may cause a discolored pixel or distorted millisecond of audio; however, in

IFTP-W over wireless networks

We simulated the performance of IFTP-W in comparison to UDP, TCP and IFTP using the same simulation configuration as in Section 5, this time adding an IFTP-W source operating simultaneously with the other three sources. All IFTP-W packets had an error-sensitive header of 40 bytes and an error-insensitive payload. Results are averaged from 10 runs of simulated connections running for approximately 2 min for each value of BER.

Effect of error-sensitive length of IFTP-W

To determine the performance of video and audio applications using IFTP-W on wireless links we ran several simulations. Our network topology consisted of three data sources and three data sinks connected through a FCFS bottleneck router as illustrated in Fig. 1. Each source operated on a different protocol – one simulated a UDP stream, one simulated a TCP connection, and the third simulated an IFTP-W stream. The default packet size for packets containing streaming video data over EDCF on

Rationale

Recently, there are four proposed TCP protocols optimized for wireless networks and based on the end-to-end paradigm. They are JTCP [19] and TCP-Jersey [20], TCP Westwood [21] and TCP-Veno [9]. JTCP uses the inter-arrival jitter and the jitter ratio to infer network congestion. While TCP-Jersey estimates the congestion window after the loss has occurred using the available bandwidth estimator algorithm. TCP Westwood frees TCP from the traditional AIMD algorithm and rather statistically

Conclusions

In this paper we proposed two new end-to-end protocols for the transmission of multimedia applications on heterogeneous wired-wireless networks. The first is IFTP-W which partitions packets into error-sensitive and error-insensitive sections. By only discarding packets with errors in the error-sensitive section, both goodput and throughput are increased. The length of the error-sensitive section is an important parameter in IFTP-W. We study the effect of its variance for video and audio packets

References (26)

  • H. ElAarag et al.

    Performance evaluation of TCP connections in ideal and non-ideal network environments

    Computer Communications Journal

    (2001)
  • D. Chiu et al.

    Analysis of the increase/decrease algorithms for congestion avoidance in computer networks

    Journal of Computer Networks and ISDN

    (1989)
  • S. Floyd et al.

    Promoting the use of end-to-end congestion control in the Internet

    IEEE/ACM Transactions on Networking

    (1999)
  • H. ElAarag et al.

    An internet friendly transport protocol for continuous media over best effort networks

    International Journal of Communication Systems

    (2002)
  • IETF, User Datagram Protocol, RFC 768, August 1980,...
  • IETF, Transmission Control Protocol, RFC 793, September 1981....
  • IETF, TCP Slow Start, Congestion Avoidance, Fast Retransmit, and Fast Recovery Algorithms, RFC 2001, January 1997,...
  • J. Widmer et al.

    A survey on TCP-friendly congestion control

    IEEE Network

    (2001)
  • M. Handley, S. Floyd, J. Pahdye, J. Widmer, TCP Friendly Rate Control (TFRC): Protocol Specification,...
  • H. ElAarag

    Improving TCP performance over mobile networks

    Journal of ACM Computing Surveys

    (2002)
  • C.P. Fu et al.

    TCP Veno: TCP enhancements for transmission over wireless access networks

    IEEE Journal on Selected Areas in Communication

    (2003)
  • S. Zabir

    Ensuring fairness among ECN and non-ECN TCP over the Internet

    International Journal of Network Management

    (2003)
  • L. Larzon, M. Degermark, S. Pink, Efficient Use of Wireless Bandwidth for Multimedia Applications, in: IEEE MoMUC,...
  • Cited by (4)

    View full text