Improving the performance of the web proxy server through group prefetching

Abstract

The web proxy server has widely been used to reduce the internet latency that the client perceived. With the help of the proxy server, the client may receive requested web objects from it rather than from the web server hosting them. Nevertheless, the client still needs to spend the time waiting for the web object being transferred from the proxy server when the local web browser does not have a valid copy of the requested object. This period of latency could be further reduced by predicting what web objects the client may need in the near future and then prefetching them to the cache of the local web browser. Different approaches had been proposed to help the proxy server capable of making prediction and prefetching to further reduce the Internet latency. Most of them use techniques including temporal locality, data mining, or more complicated mathematic models such as the Markov model to devise different predicting algorithms. We propose a new model, which dynamically combines temporal and spacial locality to help the proxy server make web object prediction and prefetching. Through the simulation against real traces, our model shows that with our design, collectively, the local web browser can lift its caching performance by an increase of 30% to 40%.