Prefetching is one of the most popular techniques for dealing with the slow access speed of the World Wide Web. To provide a mobile user with effective real time online prefetching requires that the prefetch decision is able to adapt to different network systems. This article describes an adaptive network prefetch scheme which accomplishes this task. The basic scheme is comprised of a prediction module and a threshold module, which computes the access probabilities and prefetch thresholds respectively. The access probabilities indicate how likely files will be requested by the user, and the prefetch thresholds determine whether the performance may be improved by prefetching certain files. As a user changes network in a mobile environment, it is the prefetch threshold, which is computed based on system conditions as well as the costs of bandwidth and time, that adjusts the number of prefetched files accordingly. In addition, by extending the method of computing the access probabilities, we are able to prefetch a group of files together for a user who is about to be disconnected from the network.
展开▼