Disk Space Consumption by Triple Storage Systems

Abstract

AbstractResource Description Framework (RDF) is a widespread standard and flexible method of data representation. RDF storage systems are actively used for storing, sharing, and publishing RDF data on the Internet. RDF models are used in business applications and by research teams who wish to share their data with the community. Generally, most RDF stores are optimized for queries, but usually, this is achieved at the cost of increased disk space consumption. Renting a dedicated server with a large volume of local storage is quite expensive, especially for small research teams and business startups which makes it an important factor in choosing the data storage. In this study we compared disk space usage of four popular triple storage which can serve as SPARQL (an SQL-like query language) endpoints, depending on the amount and structure of the loaded RDF data. To the best of our knowledge, no previous work has compared the disk space occupied by triple stores. We found that all of the compared open-source solutions, namely Apache Jena Fuseki, consume large amounts of hard disk space and should be used with caution in resource-limited environments. The data structure – one large graph or many smaller named graphs – strongly affected Parliament’s disk space usage so it also should be taken into account when selecting an RDF storage. Free versions of commercial systems show adequate disk consumption and appear to be weakly dependent on data structure, but Ontotext GraphDb is deliberately limited in performance, and Stardog is limited in license term and may need additional manual maintenance.KeywordsRDFSPARQLTriple storeFile storage