Hugh Smith

Overcoming the effects of correlation in packet delay measurements using inter-packet gaps

The end-to-end delay of packets in data streams is characterized with emphasis on effects due to cross traffic, sending rate and packet size. Measurements indicate that modeling delay of a packet stream with high sending rates, as a fraction of bandwidth, is difficult due to the correlations among the delay values. The correlations among inter-packet gaps (IPG) at these rates, however, are negligible. At lower sending rates, the delay correlations are negligible and the distribution of delay can be used as a delay model. We exploit the relation between delay and IPG to show that end-to-end delay can be approximated by a Markov process. Thus, a complete solution is presented for modeling delay for all sending rates. Further, a correlation estimation model is provided for delay and IPG values.

Bandwidth allocation for layered multicasted video

Some type of transmission rate control protocol is required in order to support multicast video applications over the Internet. Previously, researchers have proposed two basic families of control protocols. The first, sender based rate-adaptation, focuses on the sender. The sender uses a single channel to transmit the video signal to all receivers, and the sender adjusts the transmission rate of this single channel based upon network or receiver feedback. The second protocol focuses on the receivers. The sender stripes the video signal across multiple multicast channels, and each receiver selectively adds and drops channels to meet their individual needs. We propose a new protocol Layered Multicast Control Protocol (LMCP), which melds the strengths of the two previous approaches. The sender stripes the video signal across multiple multicast channels, and each receiver selectively adds and drops channels to meet their individual needs. The receivers also send feedback to the sender, and the sender uses this feedback to adjust the transmission rate for each channel. We present and analyze three different algorithms the sender might use for computing the transmission rates for each channel. Our analytical and simulation results show that this protocol achieves essentially optimal bandwidth utilization at the receivers.

Pattern smoothing compressed video transmission

In this paper we introduce a video smoothing algorithm for MPEG compressed live video. This algorithm, called pattern smoothing, transmits compressed video via both constant bit rate (CBR) and variable bit rate (VBR) channels. In order to take advantage of the gains achieved through statistical multiplexing of multiple sources over a single link, this algorithm utilizes a CBR channel to reduce the peak rate and variance of the VBR transmission. In addition to presenting this new algorithm, we compare it against three smoothing techniques presented in the literature. Key attributes used for comparison include receiver buffer size, live video support, startup delay, losslessness versus lossiness, and smoothing scale. Because network utilization is the most important performance metric for any smoothing algorithm, we provide a performance analysis of the pattern smoothing algorithm via simulation and compare these results to the best of the three presented smoothing algorithms.

Modelling, evaluation, and adaptive control of an instrumentation system

We present results from modeling and evaluating the JEWEL instrumentation system (IS), which is being used for runtime data collection from a distributed, real-time application. Our modeling and evaluation effort addresses two objectives: (1) providing early feedback to the system developers regarding the JEWEL IS configuration options for this application; and (2) evaluation of the design alternatives for an adaptive controller to control the overhead and intrusion of the JEWEL IS to a real-time video conferencing application. For JEWEL IS design, we compare two data collection and forwarding policies (collect-and-forward and batch-and-forward). For the design of the adaptive controller, we compare two adaptation policies (static and dynamic adaptation) and two policies to schedule the implementation of the control decisions (distributed and centralized scheduling). Results reported in this paper indicate that the batch-and-forward policy for IS design static adaptation policy with distributed scheduling for the adaptive controller design meet the domain-specific requirements.

Fair link sharing with layered multicast videoconferencing

Some type of transmission rate control protocol is required in order to support multicast video applications over the Internet. Previously, we proposed a protocol, the layered multicast control protocol (LMCP), which utilizes both the sender and receiver to control the rate of the video transmission. The sender stripes the video signal across multiple multicast channels, and each receiver selectively adds and drops channels to meet their individual needs. The receivers also send feedback to the sender, and the sender uses this feedback to adjust the transmission rate for each channel. One weakness of this approach is that it did not share the available bandwidth fairly among multiple video applications. In this paper we introduce a new router based approach for determining a video applications fair share of the available network bandwidth. This approach builds on the strengths of the LMCP approach while allowing multiple video applications to share the available bandwidth. This approach requires low overhead on the network routers, scales well to hundreds of video sources and is independent of the number of video receivers. We show through our simulation results that this approach allows the network to be shared fairly among video applications and adjusts gracefully to fluctuations in available bandwidth.

Call admission control for video phone applications

The increase in network capacities has led to the feasibility of real-time video applications such as video telephony. Such applications can use enormous amounts of bandwidth and must be carefully managed to avoid depleting network resources. One such bandwidth management approach utilizes router support to effectively control the video bandwidth usage [Hugh M. Smith and Matt W. Mutka, November 2000]. This protocol allows real-time video applications to fairly share network links and limits the bandwidth used by these applications without causing packet loss. However, this basic control algorithm does not include a mechanism to provide call admission control (CAC). This paper proposes two statistical based CAC algorithms. These algorithms attempt to minimize the impact of the CAC on the users video quality, network utilization and router processing while providing dynamic bandwidth control without packet loss. We provide simulation results of these two algorithms and compare their performance with respect to bandwidth utilization, number of calls rejected, and scalability. These simulations show that our algorithms effectively control the admission of new video sessions.

Feedback Scalability for Multicast Videoconferencing

Some type of transmission rate control is required in order to support multicast video applications over the Internet. Previously, we proposed a new protocol, the Layered Multicast Control Protocol (LMCP), which utilizes both the video sender and the receivers to control the rate of the video transmission. One weakness of this approach is that feedback from all receivers is required in order for the video source to determine an optimal transmission rate.In this paper we introduce an algorithm that allows us to determine the video sources transmission rates based on feedback from a subset or sample of the receivers feedback. As our analysis shows this algorithm allows us to support hundreds of receivers and allows the sender to determine transmission rates, which are nearly optimal. Based on the distributions analyzed, we were able to calculate transmission rates that achieved a displayable video rate within 5% of the optimal setting for 500 receivers with a worse case feedback rate at the source of 2kbps.