The Internet's excellent scalability and robustness result in part from the end-to-end nature of Internet congestion control. End-to-end congestion control algorithms alone, however, are unable to prevent the congestion collapse and unfairness created by applications that are unresponsive to network congestion. To address these maladies, we propose and investigate a novel congestion-avoidance mechanism called network border patrol (NBP). NBP entails the exchange of feedback between routers at the borders of a network in order to detect and restrict unresponsive traffic flows before they enter the network, thereby preventing congestion within the network. Moreover, NBP is complemented with the proposed enhanced core-stateless fair queueing (ECSFQ) mechanism, which provides fair bandwidth allocations to competing flows. Both NBP and ECSFQ are compliant with the Internet philosophy of pushing complexity toward the edges of the network whenever possible. Simulation results show that NBP effectively eliminates congestion collapse and that, when combined with ECSFQ, approximately max-min fair bandwidth allocations can be achieved for competing flows.

Providing strict bandwidth guarantees to packet flows in the Internet is an inherently challenging task. It requires signaling mechanisms, policing mechanisms, accurate and rapid accounting of network resources and call admission control. We present a novel protocol that provides bandwidth guarantees to Internet packet flows. This protocol, called the edge-assisted quality of service (EQOS) protocol, requires modifications to only a subset of an administrative domain's routers, namely the edge routers on the domain's periphery. Legacy routers within the core of the domain require no modifications. Our protocol maintains a high network utilization by dynamic sharing of bandwidth between the reserved and best effort flows. We present the results of several simulation experiments showing that EQOS is able to provide bandwidth guarantees to competing flows.

The end-to-end nature of Internet congestion control is an important factor in its scalability and robustness. However, end-to-end congestion control algorithms alone are incapable of preventing the congestion collapse and unfair bandwidth allocations created by applications which are unresponsive to network congestion. In this paper, we propose and investigate a new congestion avoidance mechanism called Network Border Patrol (NBP). NBP relies on the exchange of feedback between routers at the borders of a network in order to detect and restrict unresponsive traffic flows before they enter the network. The NBP mechanism is compliant with the Internet philosophy of pushing complexity toward the edges of the network whenever possible. Simulation results show that NBP effectively eliminates congestion collapse, and that, when combined with fair queueing, NBP achieves approximately max-min fair bandwidth allocations for competing network flows.

Layered transmission of data is often recommended as a solution to the problem of varying bandwidth constraints in multicast video applications. Multilayered encoding, however, is not sufficient to provide high video quality and high network utilization, since bandwidth constraints frequently change over time. Adaptive techniques capable of adjusting the rates of video layers are required to maximize video quality and network utilization. We define a class of algorithms known as source-adaptive multilayered multicast (SAMM) algorithms. In SAMM algorithms, the source uses congestion feedback to adjust the number of generated layers and the bit rate of each layer. We contrast two specific SAMM algorithms: an end-to-end algorithm, in which only end systems monitor available bandwidth and report the amount of available bandwidth to the source, and a network-based algorithm, in which intermediate nodes also monitor and report available bandwidth. Using simulations that incorporate multilayered video codecs, we demonstrate that SAMM algorithms can exhibit better scalability and responsiveness to congestion than algorithms that are not source-adaptive. We also study the performance trade-offs between end-to-end and network-based SAMM algorithms.

Network architectures that can efficiently transport high quality, multicast video are rapidly becoming a basic requirement of emerging multimedia applications. The main problem complicating multicast video transport is variation in network bandwidth constraints. An attractive solution to this problem is to use an adaptive, multi-layered video encoding mechanism. We consider two such mechanisms for the support of video multicast; one is a rate-based mechanism that relies on explicit rate congestion feedback from the network, and the other is a credit-based mechanism that relies on hop-by-hop congestion feedback. The responsiveness, bandwidth utilization, scalability and fairness of the two mechanisms are evaluated through simulations. Results suggest that while the two mechanisms exhibit performance trade-offs, both are capable of providing a high quality video service in the presence of varying bandwidth constraints.

While existing research shows that reactive congestion control mechanisms are capable of providing high video quality and channel utilization for point-to-point real-time video, there has been relatively little study of the reactive congestion control of point-to-multipoint video, especially in ATM networks. Problems complicating the provision of multipoint, feedback-based real-time video service include: (1) implosion of feedback returning to the source as the number of multicast destinations increases and (2) variance in the amount of available bandwidth on different branches in the multipoint connection. A new service architecture is proposed for real-time multicast video, and two multipoint feedback mechanisms to support this service are introduced and studied. The mechanisms support a minimum bandwidth guarantee and the best effort support of video traffic exceeding the minimum rate. They both rely on adaptive, multilayered coding at the video source and closed-loop feedback from the network in order to control both the high and low priority video generation rates of the video encoder. Simulation results show that the studied feedback mechanisms provide, at the minimum, a quality of video comparable to a constant bit rate (CBR) connection reserving the same amount of bandwidth. When unutilized network bandwidth becomes available, the mechanisms are capable of exploiting it to dynamically improve video quality beyond the minimum guaranteed level.