MePPM- Memory efficient prediction by partial match model for web prefetching

Abstract

The proliferation of World Wide Web and the immense growth of Internet users and services requiring high bandwidth have increased the response time of the users substantially. Thus, users often experience long latency while retrieving web objects. The popularity of web objects and web sites show a considerable spatial locality that makes it possible to predict future accesses based on the previous accessed ones. This infact has motivated the researchers to devise new prefetching techniques in web so as to reduce the user perceived latency. Most of the research works are based on the standard Prediction by Partial Match model and its derivates such as the Longest Repeating Sequence and the Popularity based model that are built into Markov predictor trees using common surfing patterns. These models require lot of memory. Hence, in this paper, memory efficient Prediction by Partial Match models based on Markov model are proposed to minimize memory usage compared to the standard Prediction models and its derivatives.