Scaling RDF with Time
The World Wide Web Consortium抯 RDF standard primarily consists of (subject,property,object) triples that specify the value that a given subject has for a given property. However, it is frequently the case that even for a.xed subject and property, the value varies with time. As a consequence, e.orts have been made to annotate RDF triples with 搗alid time?intervals. However, to date, no proposals exist for e.-cient indexing of such temporal RDF databases. It is clearly bene.cial to store RDF data in a relational DB ?however, standard relational indexes are inadequately equipped to handle RDF抯 graph structure. In this paper, we propose the tGRIN index structure that builds a specialized index for temporal RDF that is physically stored in an RDBMS. Past e.orts to store RDF in relational stores include Jena2 from HP, Sesame from OpenRDF.org, and 3store from the University of Southampton. We show that even when these e.orts are augmented with well known temporal indexes like R+ trees, SR-trees, ST-index, and MAP21, the tGRIN index exhibits superior performance. In terms of index build time, tGRIN takes two thirds or less of the time used by any other system, and it uses a comparable amount of memory and less disk space than Jena, Sesame and 3store. More importantly, tGRIN can answer queries three to six times faster for average query graph patterns and five to ten times faster for complex queries than these systems.
Resource Description Framework temporal RDF RDF indexing
Andrea Pugliese Octavian Udrea V.S. Subrahmanian
University of Calabria Rende, Italy University Of Maryland College Park, MD 20742
国际会议
第十七届国际万维网大会(the 17th International World Wide Web Conference)(WWW08)
北京
英文
2008-04-21（万方平台首次上网日期，不代表论文的发表时间）