Improving the performance of client Web object retrieval

The growth of the Internet has generated Web pages that are rich in media and that incur significant rendering latency when accessed through slow communication channels. The technique of Web-object prefetching can potentially expedite the presentation of Web pages by utilizing the current Web page's view time to acquire the Web objects of likely future Web pages. The performance of the Web object prefetcher is contingent on the predictability of future Web pages and quickly determining which Web objects to prefetch during the limited view time interval of the current Web page. The proposed Markov-Knapsack method uses an approach that combines a Multi-Markov Web-application centric prefetch model with a Knapsack Web object selector to enhance Web page rendering performance. The Markov Web page model ascertains the most likely next Web page set based on the current Web page and the Web object Knapsack selector determines the premium Web objects to request from these Web pages. The results presented in the paper show that the proposed methods can be effective in improving a Web browser cache-hit percentage while significantly lowering Web page rendering latency.