Materialized View Selection Considering the Diversity of Semantic Web Databases

Abstract

AbstractWith the extensive use of ontologies in various domains, Semantic Web Databases (\(\mathcal{SWDB}s\)) have appeared in the database landscape. Materialized views are one of the most popular optimization structures in advanced databases. Queries represent the most important input of the problem of selecting materialized views. In the context of \(\mathcal{SWDB}\), queries are expressed using the SPARQL language. A SPARQL query consists of a set of triple patterns executed on a set of triples representing the logical level of the \(\mathcal{SWDB}\). But a \(\mathcal{SWDB}\) may have several deployments according to the used storage layout (vertical, horizontal, binary). As a consequence the process of selecting materialized views has to consider this diversity. In this paper, we first present the difficulty of the process of materializing views in the context of \(\mathcal{SWDB}\) considering the diversity of storage layouts. Secondly, we define two approaches to select materialized views. The first approach hides the implementation aspects and views are selected at the ontological level using a rule-based approach. In the second approach, views are selected at the logical level and the view selection is guided by a cost model which considers the diverse storage layouts that can be used. Finally, intensive experiments are conducted by the means of the Lehigh University Benchmark and we empirically compare our finding with state-of-the-art algorithms.KeywordsQuery TimeLogical LevelSelection NodeSPARQL QueryQuery PlanThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.