Dash: A Novel Search Engine for Database-Generated Dynamic Web Pages

Abstract

Database-generated dynamic web pages (db-pages, in short), whose contents are created on the fly by web applications and databases, are now prominent in the web. However, many of them cannot be searched by existing search engines. Accordingly, we develop a novel search engine named Dash, which stands for Db-pAge Search, to support db-page search. Dash determines db-pages possibly generated by a target web application and its database through exploring the application code and the related database content and supports keyword search on those db-pages. In this paper, we present its system design and focus on the efficiency issue. To minimize costs incurred for collecting, maintaining, indexing and searching a massive number of db-pages that possibly have overlapped contents, Dash derives and indexes db-page fragments in place of db-pages. Each db-page fragment carries a disjointed part of a db-page. To efficiently compute and index db-page fragments from huge datasets, Dash is equipped with MapReduce based algorithms for database crawling and db-page fragment indexing. Besides, Dash has a top-k search algorithm that can efficiently assemble db-page fragments into db-pages relevant to search keywords and return the k most relevant ones. The performance of Dash is evaluated via extensive experimentation.