As the number of Web users and the diversity of Web applications continues to explode, Web Quality of Service (QoS) is an increasingly critical issue in the domain of e-Commerce [re]. This paper presents experiments designed to estimate users' tolerance of QoS in the context of e-commerce. In addition to objective measures we discuss contextual factors that influence these thresholds and show how users' conceptual models of Web tasks affect their expectations. We then show how user thresholds of tolerance can be taken into account when designing web servers. This integration of user requirements for QoS into systems design is ultimately of benefit to all stakeholders in the design of Internet services.

(WWW) basically comes down to locating an appropriate Web site and to retrieving relevant information from that site. This study examined the effect of a user’s WWW experience on both phases of the search process. Twenty-five students from two schools for Dutch preuniversity education were observed while performing three search tasks. The results indicate that subjects with WWW-experience are more proficient in locating Web sites than are novice WWW-users. The observed differences were ascribed to the experts ’ superior skills in operating Web search engines. However, on tasks that required subjects to locate information on specific Web sites, the performance of experienced and novice users was equivalent—a result that is in line with hypertext research. Based on these findings, implications for training and supporting students in searching for information on the WWW are identified. Finally, the role of the subjects ’ level of domain expertise is discussed and directions for future research are proposed.

Web page loading speed continues to vex users, even as broadband adoption continues to increase. Several studies have addressed delays both in the context of web sites as well as interactive corporate systems, and a wide range of “rules of thumb ” have been recommended. Some studies conclude that response times should be allowed to grow to no greater than 2 seconds while other studies caution on delays of 12 seconds or more. One of the strongest conclusions had been that complex tasks seemed to allow longer response times. This study examined delay times of 0, 2, 4, 6, 8, 10, and 12 seconds using 196 undergraduate students in an experiment. The subjects were randomly assigned a constant delay time and were asked to complete 9 search tasks, exploring a familiar and an unfamiliar site. Plots of the dependent variables performance, attitudes, and behavioral intentions, along those delays, suggested the use of non-linear regression, and the explained variance was in the neighborhood of 2%, 5%, and 7%, respectively. Focusing only on the familiar site, explained variance in attitudes and behavioral intentions grew to about 16%. A sensitivity analysis implies that decreases in

User-perceived retrieval latencies in the World Wide Web can be improved by pre-loading a local cache with resources likely to be accessed. A user requesting content that can be served by the cache is able to avoid the delays inherent in the Web, such as congested networks and slow servers. The difficulty, then, is to determine what content to prefetch into the cache. This work explores machine learning algorithms for user sequence prediction, both in general and specifically for sequences of Web requests. We also consider information retrieval techniques to allow the use of the content of Web pages to help predict future requests. Although history-based mechanisms can provide strong performance in predicting future requests, performance can be improved by including predictions from additional sources. While past researchers have used a variety of techniques for evaluating caching algorithms and systems, most of those methods were not applicable to the evaluation of prefetching algorithms or systems. Therefore, two new mechanisms for evaluation are introduced. The first is a detailed trace-based simulator, built from scratch,

Abstract. Dynamic Web content is increasing in popularity and, by its nature, is harder to scale than static content. As a result, dynamic Web content delivery degrades more rapidly than static content under similar client request rates. Many techniques have been explored for effectively handling heavy Web request traffic. In this paper, we concentrate on dynamic content degradation, believing that it offers a good balance between minimising total cost of ownership and maximising scalability. We describe an algorithm for dynamic content degradation that is easily implemented on top of existing mainstream Web application architectures. The algorithm is based on measuring the elapsed time of content generation. We demonstrate the algorithm’s adaptability against two traffic request patterns, and explore behavioural changes when varying the algorithm’s key parameters. We find our elapsed time based algorithm is better at recognising when the server is unloaded, that the supporting architecture limits the effectiveness of the algorithm and and that the algorithm must be configured pessimistically for best results under load. 1

Abstract. There are many ways of achieving scalable dynamic web content. In previous work we have focused on dynamic content degradation using a standard architecture and a design-time “Just In Case ” methodology. In this paper, we address certain shortcomings witnessed in our previous work by using an alternate coordination based architecture, which has interesting applicability to run-time web server adaptation. We first establish the viability of using this architecture for high-volume dynamic web content generation. In doing so, we establish it’s ability to maintain high throughput in overload conditions. Then we go on to show how we used the architecture to achieve a “Just in Time ” adaptation to achieve dynamic web content degradation in a running web application server. 1

A number of factors may influence Web users' choice of which links to follow. These include assumptions about document quality and anticipated retrieval times. The present generation of World Wide Web browsers, however, provide only minimal support to assist users in making informed decisions. Web browser `link user interfaces' typically only display a document's Universal Resource Identifier (URL), whilst a simple binary colour change in the URL's anchor is used to indicate its activation history. The question then is, how do users deal with the problem of having to make such decisions when the information at hand is insufficient? We have been conducting an investigation of how users make link selections. The results show users often are forced to fall back on heuristics and improvising strategies drawn from past experience. Based upon these results, we present a prototype of the `link lens', an enhanced link designed to make such decisions easier and more productive for all users and help less experienced ones gain a better understanding of Web behaviour.

Abstract. As the number of applications that are made available over the Internet rapidly grows, providing services with adequate performance becomes an increasingly critical issue. The performance requirements of the new applications span from few milliseconds to hundreds of seconds. In spite of the continuous technological improvement (e.g., faster servers and clients, multi-threaded browsers supporting several simultaneous and persistent TCP connections, access to the network with larger bandwidth for both servers and clients), the network performance as captured by response time and throughput does not keep up and progressively degrades. Several are the causes of the poor “Quality of Web Services ” that users very often experience. The characteristics of the traffic (self-similarity and heavy-tailedness) and the widely varying resource requirements (in terms of bandwidth, size and number of downloaded objects, processor time, number of I/Os, etc.) of web requests are among the most important ones. Other factors refer to the architectural complexity of the network path connecting the client browser to the web server and to the protocols behavior at the different layers. In this paper we present a study of the performance of web services. The first part of the paper is devoted to the analysis of the origins of the fluctuations in web data traffic. This peculiar characteristic is one of the most important causes of the performance degradation of web applications. In the second part of the paper experimental measurements of performance indices, such as end-to-end response time, TCP connection time, transfer time, of several web applications are presented. The presence of self-similarity characteristics in the traffic measurements is shown. 1

Web page loading speed continues to vex users, even as broadband adoption continues to increase. Several studies have addressed delays both in the context of web sites as well as interactive corporate systems, and a wide range of “rules of thumb ” have been recommended. Some studies conclude that response times should be allowed to grow to no greater than 2 seconds while other studies caution on delays of 12 seconds or more. One of the strongest conclusions had been that complex tasks seemed to allow longer response times. This study examined delay times of 0, 2, 4, 6, 8, 10, and 12 seconds using 196 undergraduate students in an experiment. The subjects were randomly assigned a constant delay time and were asked to complete 9 search tasks, exploring a familiar and an unfamiliar site. Plots of the dependent variables performance, attitudes, and behavioral intentions, along those delays, suggested the use of non-linear regression, and the explained variance was in the neighborhood of 2%, 5%, and 7%, respectively. Focusing only on the familiar site, explained variance in attitudes and behavioral intentions grew to about 16%. A sensitivity analysis implies that decreases in

Abstract. Web service deployment is hampered by the possibility of sudden variations in request volumes. Mechanisms exist to enhance scalability in times of heavy load when the delivered content is static. However, web services typically involve dynamic content, delivered through application servers which may have little to no support for adapting to varying loads in order to ensure timely delivery. In this paper we discuss why scaling dynamic content delivery under load is difficult, we present a technique for controlled service degradation to achieve this scalability, and we present experimental results evaluating its benefits. 1