Towards Performance Evaluation of Semantic Databases Management Systems

Abstract

The spectacular use of ontologies generates a big amount of semantic instances. To facilitate their management, a new type of databases, called semantic databases ($\mathcal{S}\mathcal{D}\mathcal{B}$) is launched. Large panoply of these $\mathcal{S}\mathcal{D}\mathcal{B}$ exists. Three main characteristics may be used to differentiate them: (i) the storage layouts for storing instances and the ontology, (ii) ontology modeling languages, and (iii) the architecture of the target database management system (DBMS) supporting them. During the deployment phase, the database administrator (DBA) is faced to a choice problem (which $\mathcal{S}\mathcal{D}\mathcal{B}$ she/he needs to choose). In this paper, we first present in details the causes of this diversity. Based on this analysis, a generic formalization of $\mathcal{S}\mathcal{D}\mathcal{B}$ is given. To facilitate the task of the DBA, mathematical cost models are presented to evaluate the performance of each type of $\mathcal{S}\mathcal{D}\mathcal{B}$. Finally, two types of intensive experiments are conducted by considering six $\mathcal{S}\mathcal{D}\mathcal{B}$, both issued from industry and academic communities; one based on our mathematical cost models and another based on the studied semantic DBMS cost models.