| The RDF (Resource Description Framework) data model has been around for morethen a decade. Since RDF data model is flexible and convenient, it is gaining importancein computational biology, knowledge sharing and social communities, and the volume ofthe RDF data is also increasing at a great speed. Many applications require online updateswith flexible control, and newer version SPARQL (SPARQL Protocol and RDF QueryLanguage) added the support for updates, which the RDF storage system shoulded alsosupport. However, existing RDF storage system has focused on the optimization ofstorage and query performance, and largely disregarded updates.The x-TripleBit aims at providing an efficient solution for RDF data updates. First,we present a data block based transaction processing framework to handle the onlineupdate. The system decomposes each SPARQL query to smaller operations and assignsoperations to threads based on which data it is about to access. Then we introduce atwo-level trasaction scheduling method to exploit the available parallelism. And when weinsert data in chunks, we propagate the new data to all copies of data at some later time toimprove the throughout of the system. And we present a two-stage based incremental bulkloading architecture with the characteristic of RDF data and the storage of the system.The x-TripleBit shows good performance on query and update while compared withthe state of art RDF storage engine named TripleBit and RDF-3X. The experimentalresults shows that x-TripleBit increases the query processing by about30%than TripleBitwith some not particularly simple queries. And the performance of incremental bulkloading has been improved at about40%compared with RDF-3X while at least10timeson the performance of online updates than RDF-3X and PostgreSQL in the case of highload. |
PDF Full Text Request