An adaptive network prefetch scheme

摘要

In this paper, we present an adaptive prefetch scheme for networknuse, in which we download files that will very likely be requested innthe near future, based on the user access history and the networknconditions. Our prefetch scheme consists of two parts: a predictionnmodule and a threshold module. In the prediction module, we estimate thenprobability with which each file will be requested in the near future.nIn the threshold module, we compute the prefetch threshold for eachnrelated server, the idea being that the access probability is comparednto the prefetch threshold. An important contribution of this paper isnthat we derive a formula for the prefetch threshold to determine itsnvalue dynamically based on system load, capacity, and the cost of timenand system resources to the user. We also show that by prefetching thosenfiles whose access probability is greater than or equal to its server'snprefetch threshold, a lower average cost can always be achieved. As annexample, we present a prediction algorithm for web browsing. Simulationsnof this prediction algorithm show that, by using access information fromnthe client, we can achieve high successful prediction rates, while usingnthat from the server generally results in more hits