We discuss a source model for VBR video traffic, and present an application to dynamic bandwidth allocation of the model. A two-state Markov chain model is used to extract useful parameters for adaptively allocating transmission bandwidth to variable-bit-rate traffic. The performance of our dynamic bandwidth allocation scheme has been compared with other schemes for both a homogeneous and a heterogeneous mix of traffic. Simulation results show that our dynamic bandwidth allocation scheme outperforms the static allocation or the low-pass filter scheme.

Dynamic bandwidth allocation using adaptive prediction can significantly improve the efficiency and QoS guarantees in transporting VBR video over an ATM network. Conventionally, the time-domain least-mean-square (LMS) predictor is used, with the drawback of slow convergence. In VBR video traffic characterized by frequent scene changes, this slow convergence may result in extended periods of intractability and excessive cell loss during scene changes. We propose an adaptive wavelet predictor for dynamic bandwidth allocation. The wavelet predictor converges faster and hence, tracks scene changes better. Our simulation results show that, in comparison with the LMS predictor, the wavelet predictor reduces the prediction error by an average of 11% over the six half-an-hour-long empirical MPEG-1 traces. The dynamic bandwidth allocation using wavelet prediction significantly reduces the cell-loss-rate over various network settings, especially at large buffer sizes.

It is anticipated that video-on-demand and video as a component of multimedia services will be popular services provided by cable television and telephone companies. The proposed wide area network protocol for these services is the asynchronous transfer mode (ATM) based broadband integrated services digital network (BISDN). In order to provide research useful for this video networking effort, this paper details two protocols for packaging and reassembling Moving Picture Experts Group (MPEG) formatted video into and out of ATM cells. These protocol are then compared and contrasted in terms of: minimal generation of overhead; minimal system delay per cell; and best error recovery/concealment facilitation in the event of cell loss.

We describe a large self routing ATM switch with moderate circuit complexity which delivers packets in sequence with a small constant delay. The switch takes advantage of trunk grouping to increase throughput and has same topology as that of a banyan network. By using modules that look at k inputs (k>16) at a time and recursively dividing the packets into two groups, the proposed switch achieved a load throughput value that is relatively independent of switch size. Cell loss on the order of 10/sup -10/ can be obtained using the switch. The switch has O(nlog/sub 2/n) complexity in the number of modules where n is the number of inputs. The modules of the switch have very regular interconnections, and hence are suitable for dense VLSI implementation.

Flexibility and efficiency of multipoint, multimedia communications will be greatly enhance if participants get delivery of only the streams, each at a specified level of quality, each user wishes to receive. The authors investigate the problem of optimal distribution of multiple streams from a source to a set of destinations, subject to the destinations' requirements and the limitations on available link capacities. Users' requests are expressed by means of bids, which are non-negative values assigned by each destination to each offered stream. Assuming that the source gains an amount equal to a user's bid whenever a stream is delivered to that user, we seek distributions that maximize the sources' gain. The optimal assignment of streams to the links of a multicast tree is formulated as an integer programming problem, and an algorithm for determining the optimal assignment is presented. The algorithm consists of three procedures: listing candidate assignments on a single link, listing candidate assignments on tree paths, and evaluating the gain for each assignment. The algorithm's correctness is proved, and its computational complexity is ascertained.>

The authors propose MaTPi, a new media-access protocol for the efficient utilization of resources in a local-area optical WDMA network. Most of the earlier work on reservation based media access protocols for single-hop optical networks has assumed the availability of state-of-the-art transmitters and/or receivers which are capable of tuning to the required wavelengths within nanoseconds. Their protocol operates in the region where the tuning times of the end-devices are of the same order of magnitude as the transmission time of a packet. In the basic MaTPi protocol, network throughput is improved by effectively overlapping the tuning duration of a transmitter in a node with the transmitting times of lasers in other nodes. They extend this basic protocol to the mult-MaTPi protocol, where each node has more than one transmitter to further improve the bandwidth efficiency of the network. Preliminary simulation results indicate that this practical protocol is indeed capable of achieving high throughput with low-cost, off-the-shelf components. A number of extensions to this protocol are also discussed.>

A connection-oriented congestion control protocol for broadband ISDN is presented. The protocol employs the traffic viewpoint hierarchical design approach and is implemented by a two-layer scheme at the call and cell layers. The call layer handles call admission and provides shortest path connection routing when a call is admitted, while the cell layer allocates switch input buffer space for each traffic type according to each type's cell loss probability grade-of-service requirement. There is a direct interaction between the two layers in processing a connection request. A discrete-time queuing system with geometrically distributed service time and state-dependent Markov modulated Bernoulli process arrivals is established to model heterogeneous networking environments for this protocol. An analytical solution is developed for this queuing system. Numerical results show that the protocol can reduce network congestion as input traffic rates approach switch capacity, and increase network utilization relative to that possible in the absence of congestion control.>

The authors examine an optimal bandwidth allocation scheme for multichannel switching systems to switch trunk groups. This multichannel switching system features a simple architecture, high throughput, high channel utilization, and other properties such as preserving packet sequence, being robust to unbalanced input traffic loads, providing fair access to network transmission resources, and fast packet processing via a pipeline structure. Numerical results show that the multichannel system can achieve a throughput of 91% per channel for a trunk group of 32 channels while also providing very low average cell delay performance in the presence of correlated inputs.>

Dilated multistage interconnection network (MIN) architectures are examined, which are characterized by multiple physical links for every logical link within the network. Their performance in a packet-switching environment is studied. The multitude of links in a dilated MIN increase throughput and capacity, offer versatility in switch configuration, and provide a ready foundation for fault-tolerant switch design. Two new configurations are introduced: a dilated extra-stage cube and a cross-connected replicated-dilated network. The latter architecture is a convenient one for single-board, modular design of larger packet switches.>

A high-performance copy network, which supports multicast fast packet switching in the broadband integrated services digital network (B-ISDN), is proposed. The network uses a banyan multistage interconnection network (MIN) with speedup and requires no intermediate buffering. It has been recognized that the copying process is, in general, a difficult one. Two important design issues, namely, obtaining multiple copies of packets and assigning destination addresses of the copies, are pursued. A solution to these issues by developing a novel two-phase copying process is presented. An analytical performance model with a special structure of Markov chains, called a branching process, is formulated. Numerical results using this model for a constant number of copies as well as for a geometrically distributed number of copies show that the proposed copy network achieves very low packet blocking probability. Simulation results are provided to confirm the analytical model.>

A fast packet switching module for use in B-ISDN switching architectures is described. The module is built around a self-routing banyan network which employs dilations and replications to enhance throughput. The differing requirements of delay and loss sensitivity for various traffic types are handled by assigning priorities to packets. A new result in the form of the throughput equation of lower priority packets in a banyan network which is improved by dilation, replication and internal speed-up is presented. The delay versus throughput analysis shows that the switch module is capable of handling integrated voice-data-video traffic efficiently.>

A high-speed, integrated service switch built around a banyan multistage interconnection network (MIN) with 4*4 elements is proposed and studied. The architecture is similar to that of the fast-packet/circuit (FPC) switch. It employs n bit-parallel lines in the MIN to reduce contention and is capable of supporting integrated voice-data-video communications. The tagged-job method is used to formulate a model for the 4*4 switching element, and it is noted that this technique extends directly to the analysis of larger n*n switching elements. Performance evaluation of the switch provides throughput and delay results for fast packet switching which supports voice and data and blocking-and-loss probability behavior for circuit-switched (CS) circuit setup which supports real-time traffic such as fixed-rate video. A performance comparison shows that this switch has a higher maximum throughput than but similar delay behavior to the FPC switch for packet switching. However, its blocking-and-loss probability for CS circuit setup is significantly smaller under identical traffic conditions.>

The authors investigate the delay-throughput performance of this protocol operating under either equal or nonuniform load conditions based on the fairness criterion of equal average packet delay. Because of the complexity of the system, an approximate analytical model for selecting the p/sub i/ for this fairness criterion is developed. The p/sub i/ that satisfy the fairness criterion are determined from a Markov chain model for the protocol. The authors begin by considering the finite-buffer case and then proceed to the infinite-buffer one. They find that the performance of the protocol is insensitive to bus length, data rate, and the number of stations on the network. Hence, the network is suitable for long-distance, high-bandwidth applications as in fiber-optic local and metropolitan area systems. Simulation studies which serve to establish the range of validity of the analytical model are presented.>

The authors extend the p/sub i/-persistent protocol previously proposed to include voice for multiaccess communication over unidirectional broadcast bus networks. They use a framed approach for integrating these two traffic types. Further, they use not only speech detectors for modeling speech sources, but also a movable-boundary scheme for their protocol in which the voice subframe size is estimated from the number of voice packets transmitted in the previous frame. The authors determine the fraction of speech loss and the fraction of wasted allocated bandwidth due to the estimator by formulating a Markov chain for the number of ready voice stations at the frame boundaries. To analyze data performance, a nonhomogeneous Markov chain is constructed for a data station's buffer content just before the data slots visit that station. Then, these models are converted to homogeneous chains from which the authors determine the optimum p/sub i/ for each data slot.>

A class of distributed retransmission controls based on stochastic approximation techniques is developed for the infinite-population, slotted, nonpersistent, and virtual-time CSMA channels. These controls, which permit each terminal to adjust its retransmission probability in a simple, recursive manner as a function of channel feedback, are stable in terms of the geometric ergodicity of an imbedded Markov chain model for each channel. Simulation results demonstrate the effectiveness of the controls in supporting stable channel operation at maximum achievable throughput, recovering from an abrupt arrival of a larger number of new packets, and providing delay versus throughput performance that closely approximates that of an idealized channel in which each terminal has perfect knowledge of the number of blocked (collided) terminals.>

A protocol for multiaccess communication over unidirectional bus networks is proposed, and its performance capabilities are determined. Under this protocol, time is slotted with a slot equaling a packet's transmission time. A station with a packet to send persists in transmitting its packet in an empty slot with probability p/sub i/ until it is successful. Three criteria for fairness in selection of the p/sub i/ are modeled using Markov chains, which are solved to obtain the proper p/sub i/ that satisfy each fairness criterion. Unlike previous studies of unidirectional bus networks, stations are allowed to buffer more than one packet. The average packet delay for this protocol is bounded, and the maximum achievable throughput approaches unity with increasing buffer size. Further, the protocol provides better delay versus throughput behavior for fixed packet lengths than previous round-robin schemes, its performance is insensitive to bus characteristics, and it appears to be particularly well suited for fiber-optic network applications requiring long distances and high bandwidths. Simulation results that confirm the predicted performance are included.>

A previously proposed protocol is extended to include voice for multiaccess communication over unidirectional broadcast bus networks. The authors use a framed approach for integrating these two traffic types where a frame, which is repeated, consists of a voice subframe followed by a data subframe. They use speech detectors for modeling speech sources in order to prevent the loss of about 60% of the bandwidth that would otherwise be required, and they use a moveable-boundary scheme in which the voice subframe size is estimated from the number of voice packets transmitted in the previous frame. The authors determine the fraction of speech loss and the fraction of wasted allocated bandwidth due to the estimator by formulating Markov chain for the number of ready voice stations at the frame boundaries. To analyze data performance, they use a nonhomogeneous Markov chain or a data station's buffer content just before the data slots visit that station. By proper control of the estimator function, they can guarantee an upper bound on the average speech loss at any voice station, and a combined voice-data throughput close to unity is achieved. Results from the analysis and from a simulation match closely.>

A new multiaccess protocol is proposed for an integrated voice/data application. The protocol, which is a variation of virtual time CSMA (VT-CSMA), takes advantage of the periodicity of voice packets and possesses a number of important features. With this protocol, voice stations appear to have a dedicated time-division multiplexed (TDM) slot, and the delay of a voice packet is bounded by the length of a frame (defined to be the period between two consecutive voice packets from a voice station). Also, the amount of data added to the channel has little effect on the voice traffic. When silence detection is used, many more voice conversations can be supported without losing the dedicated-slot characteristic. This is in contrast to a movingboundary TDM system where the excessive bandwidth saved by silence detection can only be used for data. The protocol requires no global synchronization and is easy to implement. Simulation results are presented to evaluate its performance.

A comprehensive study of the stability and optimization of the infinite population, slotted, nonpersistent CSMA and CSMA/ CD channels is presented. The approach to both stability and performance optimization differs significantly from previous work, and provides a number of new results including robustness in stability and performance in the presence of channel and control parameter variations. It is first shown that both channels are unstable under the usual assumption of random retransmission delay. Pake's lemma is then applied to study the properties of a type of distributed retransmission control which provides stable channels. Basic results are in the form of inequalities which define stability regions in the space of channel and control parameters, and further permit one to specify controls which maximize channel throughput as a function of packet length and CD time with stability guaranteed. The delay versus throughput characteristic for the stabilized channels is derived and used to demonstrate the performance achievable with these channels.

Previous efforts in modeling petroleum utilization in the United States are reviewed and critiqued. A new model, based on the system dynamics approach of Forrester, is formulated and used to investigate the effects of different tax incentive and R & D funding policies on both the short- and long-term utilization of petroleum. The present model is more viable than its predecessors since it takes into account the inherent nonlinear, dynamic, feedback, and time-lag character of energy usage. Further, it permits greater flexibility in performing policy analysis and tradeoffs.

It is well known that the problem of estimating the state and parameters of discrete-time nonlinear dynamic systems from noisy measurements can be formulated as a nonlinear two-point boundary-value problem. In this paper, an algorithm is developed for solving the above boundary-value problem. The algorithm, which is based on Newton's method, requires very little storage. A detailed discussion of the algorithm is given and certain extensions of it are indicated.