Research On Tbox Replication Partition And Placement Of Web-scaled RDF Data

The rapidly growing of semantic web has made partitioning on web scaled RDFdata become an important problem in distributed RDF data management.This paper analysis the two-level structure of RDF: TBox and ABox, introducetheir different impacts on RDF data features. TBox is one of the reasons whichmaking RDF data a power-low graph. Design a partition framework according toproperties of the web scaled RDF data. A detailed description is given on how toimplement the framework. This paper proposes a tow level replacement includinggraph partition as logical partition and range partition as physical partition. Thisreplacement combines the Low communication overhead from graph partition andDynamic sharding from range partition. An experiment on subset of DBpedia whichhas more than52million triples proves that TBox-replication method can efficientlyreduce the edge cut; another BGP test shows that System with TBox-replication canimprove performance on complex SPARQL queries.Analysis and experiment show that the partition framework is good for webscaled RDF data partition, dramatically decrease the edge cut by small redundancy.And then improve the performance of SPARQL queries. The partition frameworkprovides a good data distribution management for high level computing.