Document Type


Publication Date



As web data becomes more essential in our work and play and it keeps growing in an explosive way, web view mechanisms are extensively employed to offer customized value-added services to customers and they are usually materialized to achieve fast query response time. However, the dynamicity problems of the underneath web information is not as easy to tackle as it is in the context of conventional database systems. Developing maintenance techniques for materialized web restricting the structure of all the web data sources and the shareability of web data sources enabling each update on a single data source to potentially affect many others in the web data graph. To compute web view "patches" for its incremental maintenance in response to an update, a large amount of accesses back to base data is usually inevitable, but it is obviously not desirable because of the likelihood of severe impact from the heavy network overhead the the intense contention for base data. In this paper, given a web view specification defined over a hierarchical web data graph, we analyze the query patter, conduct the evaluation strategy along aggregation paths as to distill a subgraph of web data objects, for which we set up an index structure. By utilizing the precomputed value aggregation results stored in such an index, our algorithms show that both web view computation and its maintenance can be done more efficiently. Cost analysis and experiment studies on the gains of our incremental maintenance approach compared to the state-of-art solutions are also conducted.