A Request-Based Approach to Maintain Object Consistency in Content Distribution Network

Abstract

A novel request-based object consistency algorithm for Content Distribution Network (CDN) is proposed to ensure that locally cached data is consistent with that at the server. Interesting rate is proposed to indicate access rate of object, replica server polls Web server periodically and pull changes to object whose interesting rate equals to 1. Dynamic slip-window method is used to enable adjusting the size of refresh window according to request and modify rate in previous window. In order to make refresh interval more accuracy, object request clustering algorithm is proposed to optimize the basic algorithm. Simulated results show that this algorithm can take advantage of Content Distribution Network and significantly reduce the traffic generated as well as ensure object fidelity.