Analysis of Semantic Web Databases

Abstract

The popularity of Semantic Web has given rise to the development of Semantic Web databases with improved performance. Benchmarks are being performed to validate performance claim made by developers of Semantic Web databases. However, detailed information regarding the strengths and shortcomings of these databases is limited due to the fact that the existing benchmarks provide little depth in scalability analysis. They measure the Semantic Web databases’ performance in terms of time and do not cover resource utilization during data manipulation operations. The research literature available on Semantic Web databases does not provide details of their internal architecture. In this research, we aim to evaluate the existing Semantic Web databases to discover their comparative behavior and scalability trends for a newly proposed evaluation methodology, and to analyze their architectures particularly with respect to their storage schemas and access methods. To cope with the deficiencies of existing evaluation methodologies, we have proposed a new evaluation methodology to perform comparative analysis and scalability performance study of Semantic Web databases. Our evaluation methodology comprises test cases for the data access methods and query optimization techniques to analyze the performance of Semantic Web databases. We defined new metrics for query cost estimation. As a part of this work, we also evaluated the performance of seven prominent open-source Semantic Web databases. These Semantic Web databases were evaluated on our proposed evaluation methodology using Barton Library dataset. Based upon our experiments and proposed methodology, we highlighted the key strengths and weaknesses of these Semantic Web databases, and discovered their scalability behavior. Storage schemas and access mechanism of the Semantic Web databases are identified in this thesis. We conclude that overall native Semantic Web databases perform better than others i.e. in-memory and non-memory non native Semantic Web databases. We also conclude that the requirements of in-memory stores for time and resource usage do not increase as rapidly as in other two categories of Semantic Web databases. The evaluation results show that the proposed evaluation methodology provides better scalability behavior and performance estimation of Semantic Web database than the existing evaluation studies.