Data aware hybrid partitioning technique and query execution for RDF data

Abstract

Storage of huge semantic web data using relational techniques has been an important research issue in web databases. Semantic web data is Resource Description Framework (RDF) data. Triple store technique is very inefficient because it requires higher number of self joins. To avoid this Vertical Partition Technique is used but Vertical Partition introduces more number of joins. So overcome the problem in VP, Property Table Technique is used which requires less number of joins. But property table suffers from higher null value handling problem. This thesis represents Hybrid Storage Technique which is the combination of the two techniques: Property Table and Vertical Partitioning. This work follows Data Aware method to implement proposed hybrid storage technique. This work presents Data Aware Hybrid Storage Technique for two large publicly available real-world RDF dataset (DBLP and DBPedia), for which analysis of the performance of Data Aware Hybrid Storage Technique has been done and compared with Vertical Partition Technique. This experiment does not assume any particular query workload for its structure. This work uses row-oriented relational database management tool, Postgres. For DBLP, this experiment shows Data Aware Storage Technique performs an average of 8% better than VP Technique for hot runs and 12% better for cold runs. For DBPedia, Data Aware Storage Technique performs an average of 11% better than VP Technique for hot runs and 36% better for cold runs.