This research experimentally examines the effects of network delays, document type, and various user characteristics on the perceived usability of distributed documents on the internet. Six experimental conditions were analysed: text only documents and documents including text and graphics at three levels of delay. Users were undergraduate and graduate students who reported spending zero to 50 hours per week on the internet. Usability was assessed by analysing responses to questions about ease of locating information, information organization, information quality, and navigation problems. The results showed significant interactions of network delay and document type on subjects' perceptions of quality, organization, navigation, and several additional factors that are important to organizations providing information on the internet. Both the number of hours subjects reported spending using the internet and subjects' self-reported command of the English language were significantly correlated with perceptions of various aspects of the internet sites examined. While the results indicate that internet users may prefer highly graphical web sites, it appears that they are unwilling to tolerate substantial delays. As a result, users in the study preferred plain text documents as delays increased.

We analyze 12 traces of round-trip Internet packet delay. We find that these traces, when viewed as time series data, often exhibit Hurst parameter (H) estimates greater than 0.5, indicating long-range dependence. Several traces, however, are not well-modeled with a constant H. We discuss in detail our analytical methods and the robustness of empirical estimators of H under conditions of non-negligible packet loss. Our results indicate that Internet delay is bursty across multiple time scales, which implies that end-user duality of service in the Internet is likely to be impacted by long periods of very large and/or highly variable delays.

In this paper we present results which suggest that Internet packet delay, when viewed as a time series, is self-similar in nature. A self-similar phenomenon displays the same or similar statistical properties across a wide range of time scales. Self-similarity has previously been observed in the magnitude of traffic transmitted on local-area and wide-area networks. Our research is the first to explore the self-similarity of traffic delay rather than traffic magnitude. Our results imply that the delay experienced by application layer protocols is extremely bursty and that traditional Poisson models cannot accurately predict the degree of burstiness. Furthermore, we present evidence that the degree of self-similarity for a round-trip path in the Internet may be correlated with the packet loss observed on that path.

We propose an efficient algorithm for constructing multicast trees in the Shufflenet. Although the Shufflenet has been shown to be an important network topology for wavelength division multiplexed (WDM) local optical networks, no algorithm has previously been proposed for multicasting in this network. We present a single-pass algorithm that constructs minimum-delay multicast trees, and show how a second pass can be added to reduce tree cost while maintaining very low delay.

A passive-star-based, broadcast-and-select, local lightwave network which can support a limited number of wavelength-division multiplexed (WDM) channels, but serve a much larger number of nodes, is considered. Each node is equipped with one tunable transmitter and one fixed receiver, and each WDM channel is operated in a time-division multiplexed (TDM) fashion for carrying packet traffic. Bandwidth is allocated to the node pairs when traffic flow between them is nonuniform, while also accommodating transceiver tuning latency. Our approach exploits well-known results from scheduling theory to create efficient transmission schedules. Multiprocessor task scheduling heuristics that can be applied to load balancing in a multichannel network is also examined.

Rainbow is a prototype optical metropolitan area network (MAN) developed at IBM. It employs wavelength-division multiplexing (WDM) on a fiber-optic, passive-star network topology, with each station equipped with a laser, which is fixed tuned to its own unique wavelength, and a Fabry-Perot filter, which is tunable across all wavelengths. This paper presents a model and analysis of the protocol used in the Rainbow prototype using the equilibrium point analysis (EPA) technique. We examine the system's throughput and how it is affected by various system parameters such as message arrival rate, message length, and timeout duration. We show that, for a given arrival rate, there is a timeout duration that will yield the optimal throughput. The analytical results are verified by simulation.

A passive-star-based, broadcast-and-select, local lightwave network which can support a limited number of WDM channels (on the order of ten), but serve a much larger number of nodes (a few tens or hundreds), is considered. Each node is equipped with one tunable transmitter and one fixed receiver, and each WDM channel is operated in a TDM fashion for carrying packet traffic. Bandwidth is allocated to the node pairs when traffic flow between them is non-uniform, while also accommodating transceiver tuning latency. Our approach exploits well-known results from scheduling theory to create efficient transmission schedules.