The Internet has rapidly evolved from being a scientific experiment to a commercial network connecting millions of hosts that carries traffic generated by a large amount of applications with diverse requirements. Its architecture was however designed to enable efficient point-to-point delivery of bulk data, and can not provide statistical guarantees on the timely delivery of delay sensitive data such as streaming and real-time multimedia. Thus, applications that require low loss probabilities in today's Internet have to use some end-to-end error recovery mechanism. For delay sensitive applications the introduced latency by the applied schemes has to be low as well. Traffic control functions such as delay limited shaping and forward error correction (FEC), and multiple description coding (MDC) have been proposed for variable bitrate video. Their major drawback is, however, that it is difficult to predict their efficiency, as it depends on many factors like the characteristics of the stream itself, the characteristics of the traffic in the network and the network parameters. Consequently, it is difficult to decide which control mechanisms to employ, how to combine them and to choose the right parameters (e.g. block length, code rate) for optimal performance. In this thesis we present results on the efficiency of traffic control functions and MDC for video transmission based on mathematical models and simulations. We investigate the efficiency of delay limited traffic shaping and the trade-offs in the joint use of traffic shaping and forward error correction. We identify the packet size distribution of the traffic in the network as an additional factor that may influence the efficiency of FEC, and present a thorough analysis of its possible effects. We present an analytical comparison of MDC versus media-dependent FEC and media-independent FEC, and based on the results we conclude that MDC is a promising error control solution for multimedia communications with very strict delay bounds in an environment with bursty losses. We combine the analytical results with traces from measurements performed on the Internet to evaluate how efficient these error control schemes are under real loss patterns. We compare the efficiency of MDC and media-dependent FEC in the presence of channel estimation errors; we propose a new rate allocation method, which is robust to mis-estimations of the channel state and which improves error resilience on non-stationary channels. Finally we present an analytical model of the performance of an end-point-based multimedia streaming architecture based on multiple distribution trees and forward error correction, and analyze the behavior of the architecture for a large number of nodes. / QC 20101115

Computers using the tagged-token dataflow model are among the best candidates for delivering extremely high levels of performance required in the future. Instruction scheduling in these computers is determined by associatively matching data-bearing tokens in a Waiting-Matching Unit (W-M unit). At the W-M unit, incoming tokens with matching contexts are forwarded to an instruction while non-matching tokens are stored to await their matching partner. Requirements of the W-M unit are exacting. Necessary token storage capacity at each processing element (PE) is presently estimated to be 100,000 tokens. Since the most often executed arithmetic instructions require two operands, the bandwidth of the W-M unit must be approximately twice that of the ALU. The contradictory requirements of high storage capacity and high memory bandwidth have compromised the M-W units of previous dataflow computers limiting their speed. However, tokens arriving at a PE exhibit strong temporal locality. This naturally suggests the use of some caching technique. Using a recently developed CAM memory structure as a base, a token caching scheme is described which allows rapid, fully associative token matching while allowing a large token storage capacity. The key to the caching scheme is a fast and compact, articulated, first-in, first-out, content addressable memory (AFCAM) which allows associative matching and garbage collection while maintaining temporal ordering. A new memory cell is developed as the basis for the AFCAM in an advanced CMOS (Complementary Metal Oxide Semiconductor) technology. The design of the cell is discussed as well as electrical simulation results, verifying its operation and performance. Finally, estimated system performance of a dataflow computer using the caching scheme is presented. / Graduation date: 1991

No description available.

The objective of this dissertation is to enhance the overall understanding of practical manufacturing systems by using rigorous academic approaches, primarily queueing theory. The scope spans from the performance of a single manufacturing process to the performance of a manufacturing system. Queueing models are commonly used to evaluate the performance of manufacturing systems. Exact M/M/1 or approximations of G/G/1 models are usually adopted to describe the behavior of a single machine system. However, when applying queueing models to a single machine, some practical issues are encountered. A real machine is subject to different types of interruptions, such as breakdowns, setups and routine maintenance. The proper queueing models under interruptions are presented. The behavior of manufacturing systems is explored by first investigating the underlying structure of tandem queues. We introduce two properties describing the dependence among servers in tandem queues, namely the intrinsic gap and intrinsic ratio, and develop a new approximation approach. The approach exploits what we call the nearly-linear and heavy-traffic properties of the intrinsic ratio. Across a broad range of examined cases, this new approach outperforms earlier approximations that are based on the parametric-decomposition and diffusion approximation approaches. We also demonstrate its use with historical data to achieve very accurate queue time estimates. Furthermore, based on the structure of tandem queues, a way to model the performance of manufacturing systems has been developed.

<p>The Internet has rapidly evolved from being a scientific experiment to a commercial network connecting millions of hosts that carries traffic generated by a large amount of applications with diverse requirements. Its architecture was however designed to enable efficient point-to-point delivery of bulk data, and can not provide statistical guarantees on the timely delivery of delay sensitive data such as streaming and real-time multimedia. Thus, applications that require low loss probabilities in today's Internet have to use some end-to-end error recovery mechanism. For delay sensitive applications the introduced latency by the applied schemes has to be low as well. Traffic control functions such as delay limited shaping and forward error correction (FEC), and multiple description coding (MDC) have been proposed for variable bitrate video. Their major drawback is, however, that it is difficult to predict their efficiency, as it depends on many factors like the characteristics of the stream itself, the characteristics of the traffic in the network and the network parameters. Consequently, it is difficult to decide which control mechanisms to employ, how to combine them and to choose the right parameters (e.g. block length, code rate) for optimal performance.</p><p>In this thesis we present results on the efficiency of traffic control functions and MDC for video transmission based on mathematical models and simulations. We investigate the efficiency of delay limited traffic shaping and the trade-offs in the joint use of traffic shaping and forward error correction. We identify the packet size distribution of the traffic in the network as an additional factor that may influence the efficiency of FEC, and present a thorough analysis of its possible effects. We present an analytical comparison of MDC versus media-dependent FEC and media-independent FEC, and based on the results we conclude that MDC is a promising error control solution for multimedia communications with very strict delay bounds in an environment with bursty losses. We combine the analytical results with traces from measurements performed on the Internet to evaluate how efficient these error control schemes are under real loss patterns. We compare the efficiency of MDC and media-dependent FEC in the presence of channel estimation errors; we propose a new rate allocation method, which is robust to mis-estimations of the channel state and which improves error resilience on non-stationary channels. Finally we present an analytical model of the performance of an end-point-based multimedia streaming architecture based on multiple distribution trees and forward error correction, and analyze the behavior of the architecture for a large number of nodes.</p>

Thesis (M. Phil.)--University of Hong Kong, 2001. / Includes bibliographical references (leaves 87-90).

Thesis (Ph. D.)--University of Massachusetts Amherst, 2009. / Includes bibliographical references (p. 266-272). Print copy also available.

Thesis (M.S.)--Worcester Polytechnic Institute. / Keywords: AECN; heterogeneous flows; RED; ECN; goodput; fairness. Includes bibliographical references.

published_or_final_version / Mathematics / Master / Master of Philosophy

Multiclass queueing networks are widely used to model complex manufacturing systems and communications networks. In this dissertation we describe and analyze a multiclass queueing network model known as the Cambridge Ring. As the name suggest this network has a circular topology with unidirectional routing. In many cases the analysis of a stochastic model is a difficult task. For a few special cases of this network we show that all non-idling policies are throughput optimal for this system. One of the major differences between this work and precious literature is that we prove throughput optimality of all non-idling policies, whereas most of the previous work has been on establishing throughput optimality for a specific policy (usually First-In-First-Out). We use a macroscopic technique known as fluid model to identify optimal policies with respect to work in process. In one case we consider, the discrete scheduling policy motivated by the optimal fluid policy is indeed optimal in the discrete network. For the other special case we show by means of a deterministic counterexample that the discrete policy most naturally suggested by the fluid optimal policy may not be optimal for the queueing network. We also formulate the fluid holding cost optimization problem and present its solution for a simple version of the Cambridge Ring. Further we establish that the optimal policy under a class of policies known as "non-ejective" policies may be an idling policy. We use an example of the Cambridge Ring with a single vehicle to show that the optimal policy for this example has to be an idling policy. / text