MCBDist: A Novel Markov-Chain-Based Measure of Distance among Webpages

In the Web server cluster adopting content-aware request dispatching, the persistent connection feature of HTTP/1.1 may bring extra cost. If the webpages, that are likely to be visited together by users, are grouped into webpage cluster appropriately, and the webpage cluster is viewed as document distribution unit, such cost may decrease. How to measure the distance among webpages is a key problem of webpage clustering. In this paper, we propose a novel Markov-chain-based measure of distance among webpages, called MCBDist. It not only considers the temporal correlation of users´ visit, but also considers the path of users´ visit. In addition, for the dimension of the transition probability matrix is usually very large, transition probability matrix compression is used to deduce the computational complexity. Finally, we give an example to illustrate the effectiveness of MCBDist.