Modelling dynamic content behavior for efficient WEB caching and distribution

Abstract

This thesis examines the dynamic nature of web content at both object level and fragment level. Statistical models and quantity matrixes are built for the modification dynamics of web content at different granularities. The similarities of different modification patterns are modeled and quantified. A systematic study is also carried out to understand the object decomposition process and its performance impacts on content delivery. Heuristic algorithms for cost-aware decomposition are proposed and evaluated. Preferences of the decomposition algorithms on content characteristics are investigated under different selection polices of validation frequency in simplified content management practice. Finally, a proxy-based system framework is proposed for fragment delivery with a decision-decoupled cache consistency model. This proposed framework includes system requirements such as an RDF-based markup language support, HTTP protocol extension, automatic cost-aware decomposition mechanism and selective fragment bundle mechanism. With the decision-decoupled cache consistency mechanism, the proxy-based fragment delivery framework in this thesis has the potential to efficiently deliver more dynamic and heterogeneous contents to the clients.