Elsevier

Computer Communications

Volume 22, Issue 10, 25 June 1999, Pages 930-941
Computer Communications

Transmitting real-time VBR traffic with QoS control in a timed token medium access control network

https://doi.org/10.1016/S0140-3664(99)00061-4Get rights and content

Abstract

By exploring the characteristics of the Moving Pictures Experts Group (MPEG) coded frame sequence and the expected quality of service (QoS) of video applications, we have designed different transmission strategies for transmitting real-time VBR videos over a computer network. In this article, we present three MPEG video transmission schemes with the objective of increasing the maximum number of video streams that can be supported by the underlying network without sacrificing much on the video quality. Multiple classes of the MPEG video are examined in this study. These data are captured from real video programmes and we categorized these video clips according to their traffic burstiness and workload characteristics. The results reveal that, by employing the intentional delay scheme, we could improve the performance by 67–166%. In the case of applications, allowing different levels of QoS, the QoS Control Scheme alone can improve the performance by 77 and 112% when the QoS level is set to 90 and 60%, respectively. When we combined these two schemes together and formed our third scheme, namely the QoS Control with intentional delay, the performance improvement can be pushed to 271%, which is a dramatic improvement over the default uncontrolled transmission scheme.

Introduction

As a result of the demands of the videos for network bandwidth, video data needs to be compressed before being transmitted over a computer network. Many compression schemes have been proposed or developed for these variable bit rate (VBR) video services. Of all these video compression schemes, the scheme developed by the Moving Pictures Experts Group (MPEG) is the most notable one. By the MPEG standard, video information is compressed frame by frame, and the resultant frames are transmitted and processed under stringent timing constraints. To support such applications on top of a computer network, the transmission scheme must ensure that most, if not all, frames should reach their destinations before their deadlines. The frames arriving late and missing their deadlines will result in a poorer quality of the picture or jumpy video. Hence, there is a tradeoff between video quality and on-time delivery of video frames.

In this article, we deal with real-time video traffic and not stored videos. Hence, we assume that the sending and the receiving stations have limited buffers, and that techniques for video smoothing are not applicable here. In order to address the issue of on-time video frame delivery, transmission delays for the underlying network must be deterministic and bounded. Since the aim of this article is not to find the best network protocol for the transmission of the MPEG videos, we adopted the timed token medium access control protocol here for our investigation just because of its important property of bounded transmission delay. This bounded delay property is a necessity for real-time communications, multimedia computing systems, and industrial process controls. This is why the timed token MAC protocol is well received and is being adopted in several high-bandwidth network standards over the years. These include the fiber distributed data interface (FDDI) [1], [2], IEEE 802.4 (token bus) [3], the high-speed data bus and the high-speed ring bus (HSDB/HSRB) [4], [5], [6], and the survivable adaptable fiber optic embedded network (SAFENET) [7], [8], [9]. Besides, many embedded real-time applications also used the timed token MAC protocol as their backbone network.

Furthermore, some of our previous studies [10], [11], [12], [13] have shown that the timed token MAC protocol is suitable for real-time communication and multimedia application within a local area network (LAN). The main idea behind the timed token MAC protocol is to control the sharing of the network bandwidth among the stations for transmitting synchronous messages. That is, the bandwidth allocation scheme for synchronous traffic. Among the studies on the bandwidth allocation schemes in the timed token MAC protocol [14], [15], [16], [17], [18], Agrawal, Chen, Malcolm, and Zhao [14], [15], [16], [17] proposed a normalized proportional allocation scheme, which will give the highest worst case achievable utilization and the utilization bound is at 33%. That is, as long as the total synchronous traffic is no more than 33%, these synchronous messages are guaranteed to be transmitted before their deadlines. Since this bandwidth allocation scheme is well accepted in the real-time communication community and that it is the best scheme among the others, such as, the full length allocation scheme, the proportional allocation scheme, and the equal partition allocation scheme,… etc. Hence, we adopted the normalized proportional allocation scheme for our MPEG transmission experiments. In fact, in our later studies, we found out that different bandwidth allocation schemes do not affect the conclusion of our results.

However, the VBR video encoding schemes as well as the MPEG-I video transmission over a computer network have been studied extensively. Ott et al. [19] and Lam et al. [20] proposed smoothing schemes for VBR video. Reibman and Berger [21] and Reininger et al. [22] studied the problem of transporting/multiplexing VBR/MPEG video over ATM networks. Pancha and Zarki [23], [24] studied the MPEG-I video coding standard for transmission of VBR video and the performance of variable bandwidth allocation schemes for VBR MPEG-I video. Furthermore, there were also extensive studies on MPEG video characterization and modeling by Ismail et al., Izquierdo and Reeves, and Krunz et al., [25], [26], [27].

In this article, not only did we propose three transmission schemes for the transport of the MPEG videos over a network, we also analyze and constructed a series of simulation experiments to study the performance of a timed token MAC network in supporting the real-time VBR traffic transmissions. This article proposed three schemes: the intentional delay scheme, the QoS control scheme, and the combined scheme of QoS control with intentional delay for improving the real-time support of the VBR transmission on a high-speed network with a timed token MAC protocol. What makes these experiments different from the others is that the video data we used are captured from various TV programmes. We categorized these video clips according to their workload characteristics—average frame size, maximum frame size, and burstiness specified by the I:P:B frame size ratio—and studied their effect on the transmission schemes and the underlying network. The reader should notice that by MPEG standard, we are referring to the MPEG-I standard as specified by ISO/IEC-11172. From here onwards, we will omit the suffix and use the MPEG instead of MPEG-I. However, the reader should note that our proposed transmission schemes are also applicable to MPEG-II, and other VBR video streams. In fact, the proposed transmission schemes can be generalized and applicable to not just the MPEG video traffic, but to other real-time communication applications and time critical industrial controls over a transmission delay bounded network.

The remainder of this article is organized as follows: Section 2 describes the system model and the network protocol. Section 3 describes the MPEG video streams, encoding parameters, and their classifications. Section 4 discusses the motivation, observations, and the analysis of the proposed transmission schemes. Section 5 describes the simulation experiments, performance metrics, and presents the simulation results. Finally, Section 6 provides a summary of this study.

Section snippets

System model

We consider a network with n stations connected by point-to-point links forming a ring. The medium access control is controlled by token-polling from station to station. In a network with a timed token MAC protocol, messages are classified into two categories: synchronous and asynchronous messages. Synchronous messages arrive or are ready to be transmitted at regular time intervals (periodic) and are associated with deadline constraints. In contrast, asynchronous messages are non-periodic and

MPEG video stream

The MPEG was originally designed for storing video and audio in the digital media, however, the MPEG compression is also used for transmitting video frames over a computer network. The basic idea of this compression scheme is to predict the motion from frame to frame in the temporal direction, and then to use discrete cosine transforms (DCT) to organize the redundancy in the spatial directions. In the MPEG standard, there are three types of frames forming a group of pictures (GOP). The I frames

MPEG transmission schemes

In this section, we present three transmission schemes in order to improve the transmission of the MPEG video streams over a timed token MAC network. As we are only interested in dealing with the real-time VBR streams such as those in a video conferencing, and a video supervision applications, video frames have to be a real-time encoded and ready to be sent periodically. We also assume that the sending and receiving stations have limited buffers and that video smoothing techniques to transform

Simulation experiment and results

Although for each of our proposed scheme, we derive its maximum bandwidth demand and show that all of them perform better than the default transmission scheme. However, because of the VBR in a MPEG video stream, and the number of streams per system evaluated, it is difficult to tell which transmission scheme is the best. Therefore, we have constructed a series of experiments to find out the PI among the proposed schemes as compared with the default transmission scheme.

Summary

As a summary of this performance study, we first categorized the MPEG streams into four different types according to their I:P:B frame size ratios. By examining the bandwidth demand and the transmission sequence, we proposed two ways to improve the support of a timed token MAC network for MPEG video transmissions. With the Intentional Delay Scheme, video streams are regulated such that they are not competing for the bandwidth at the same time, thus, the bandwidth can be utilized effectively.

Acknowledgements

The work reported in this paper was supported in part by the RGC Earmarked Research Grant under RGC/97-98/54, and by the FRG under FRG/96-97/II-103.

J.K.-Y. Ng received the BS degree in Mathematics and Computer Science, the MSc degree in Computer Science, and the PhD degree in Computer Science all from the University of Illinois at Urbana-Champaign in the years 1986, 1988, and 1993, respectively. Dr. Ng is currently an assistant professor in the Department of Computer Science at Hong Kong Baptist University. His research areas include: real-time networks, multimedia communication, ATM delay analysis, high-speed network simulation, and

References (31)

  • FDDI, Token ring media access control (MAC). ANSI Standard X3.139,...
  • FDDI, Token ring station management, draft proposal. ANSI Standard X3T9.5 rev. 5, 10 May...
  • IEEE/ANSI Standard 802.4-1985, Token passing bus access method and physical layer specifications, IEEE, New York,...
  • M.D. Cohn, A network architecture for advanced aircraft, in: Proceedings of the IEEE Conference on Local Computer...
  • Aerospace Systems Division, Committee AS-2, Linear token-passing multiple data bus, AS4074.1 Version 4.0, 25 January...
  • R.W. Uhlhorn

    The fibre-optic high-speed data bus for a new generation of military aircraft

    Ieee Lcs

    (1991)
  • D.T. Green, D.T. Marlow, SAFENET–a LAN for navy mission critical systems, in: Proceedings of the Fourteenth Conference...
  • R.J. Kochanski et al.

    Safenet–the standard and its application

    Ieee Lcs

    (1991)
  • J.L. Paige, SAFENET–A navy approach to computer networking, in: Proceedings of the IEEE Conference on Local Computer...
  • J.K. Ng, J.W.S. Liu, Performance of local area network protocols for hard real-time applications, in: Proceedings of...
  • J.K. Ng, The support of different network medium-access protocols for multimedia applications, in: Proceedings of the...
  • J.K. Ng, J.W.S. Liu, Performance of high-speed networks for multimedia applications, in Proceedings of the eighteenth...
  • J.K. Ng, Performance of high-speed networks for real-time applications, PhD thesis, University of Illinois at...
  • G. Agrawal et al.

    Guaranteeing synchronous message deadlines with the timed token medium access control protocol

    Ieee Transaction on Computers

    (1994)
  • B. Chen, G. Agrawal, W. Zhao, Optimal synchronous capacity allocation for hard real-time communication with the time...
  • Cited by (1)

    J.K.-Y. Ng received the BS degree in Mathematics and Computer Science, the MSc degree in Computer Science, and the PhD degree in Computer Science all from the University of Illinois at Urbana-Champaign in the years 1986, 1988, and 1993, respectively. Dr. Ng is currently an assistant professor in the Department of Computer Science at Hong Kong Baptist University. His research areas include: real-time networks, multimedia communication, ATM delay analysis, high-speed network simulation, and distributed systems performance evaluation. Dr. Ng has been a member of the IEEE and the IEEE Computer Society since 1991, and has been an exco-member (1993–1995), General Secretary (1995–1997), and Vice-Chair (1997, present) of the IEEE, Hong Kong Section, Computer Chapter. He is also a member of the ACM, EUROMICRO.

    V.C.-S. Lee received his BSc (Hons) degree in Information Technology in 1990 from City Polytechnic of Hong Kong. He received his MPhil degree in Computer Science in 1994 from the same institute and the PhD degree in Computer Science in 1997 from City University of Hong Kong. Dr. Lee is currently a research fellow in the Department of Computer Science at the City University of Hong Kong. His research interests include: high-speed networks, real-time databases and mobile computing. He is also a member of the IEEE and ACM.

    View full text