Currently there is no control for real-time traffic sources in IP networks. This is a serious problem because real-time traffic can not only congest the network but can also cause unfairness and starvation of TCP traffic. A new rate control scheme, RCS, is introduced for real-time traffic in networks with high bandwidth-delay products and high bit error rates. RCS is based on the concept of using dummy packets to probe the availability of network resources. Dummy packets are treated as low priority packets and consequently they do not affect the throughput of actual data traffic. Therefore, RCS requires all the routers in the connection path to support some priority policy. Simulation experiments show that in environments with high bandwidth-delay products and high bit error rates, RCS achieves good throughput performance without penalizing TCP connections.

Development of the Next Generation Internet requires the development of revolutionary applications as well as advances in networking technologies.This paper presents experiences in using the NREN testbed to prototype a collaborative medical-imaging application.Technological requirements for this application include high-bandwidth reliable multicast and QoS provisioning.Engineering the network infrastructure for a demonstration in May 1999 was a challenging experience.We achieved high-bandwidth multicast for the demonstration; current efforts focus on satisfying application requirements for reliable multicast and QoS.

The fiber distributed data interface (FDDI) is an ANSI draft proposed standard for a 100 Mbit/s fiber-optic token ring. The FDDI timed token access protocol provides dynamic adjustment of the load offered to the ring, with the goal of maintaining a specified token rotation time and of providing a guaranteed upper bound on time between successive arrivals of the token at a station. FDDI also provides automatic recovery when errors occur. The bound on time between successive token arrivals is guaranteed only if the token rotates quickly enough to satisfy timer requirements in each station when all ring resources are functioning properly. Otherwise, recovery would be initiated unnecessarily. The purpose of this paper is to prove that FDDI timing requirements are satisfied, i.e., the token rotates quickly enough to prevent initiation of recovery unless there is failure of a physical resource or unless the network management entity within a station initiates the recovery process.

The FDDI Token Ring Protocol controls communication over fiber optic rings with transmission rates in the range of 100 megabits per second. It is intended to give guaranteed response to time-critical messages by using a "timed token" protocol, in which non-critical messages may be transmitted only if recent movement of the token among stations has been sufficiently fast relative to a "target" token rotation time (TTRT).

Communication technology now makes it possible to support high data transmission rates at relatively low cost. In particular, optical fiber can be used as the medium in local area networks with data rates in the range of 100 megabits per second. Unfortunately, local area network topologies and communication protocols that work well with lower speed media are not necessarily appropriate when the data transmission rate is scaled up by approximately an order of magnitude. Recognizing this fact, an ANSI sub-committee (ANSIX3T9) has been working for the past two years on a proposed standard for a token ring protocol tailored to a transmission medium with transmission rate in the 100 megabits per second range. The protocol is referred to as the FDDI (Fiber Distributed Data Interface) Token Ring protocol. The proposal for the standard is now quite mature and nearly stable.

Abstract This paper is an analysis of overhead required by the MERIT routing table update algorithm after failure of a single network resource. The number of update messages sent by each node during the update process is used as the measure of update overhead. Both synchronous behavior and asynchronous behavior of the MERIT algorithm are considered. The number of messages required for synchronous update is shown to be an upper bound for the number of messages required if update occurs asynchronously. The effect of network topology on update overhead is discussed. Finally, some implementation suggestions aimed at minimizing looping of messages during routing table update are presented.

In a distributed computer network each node participates in the routing process, making routing decisions based on information about network topology which is stored in tables in the node. To maintain the accuracy of these tables without interrupting the functionality of the network, they must be dynamically updated whenever there is a topology change in the network. The purpose of this paper is to analyze update activity required by a topology information maintenance scheme which has been implemented on the three-node MERIT Computer Network in Michigan. The main theorem gives upper and lower bounds on the number of update messages required after failure of a single network resource, either a node or a link.