Just out of curiosity...has any benchmarking been performed on Jena TDB. If yes, then what were the datasets used? And what were their running times?
On Tue, Jul 30, 2013 at 5:32 PM, Andy Seaborne <[email protected]> wrote: > (lots of questions - maybe if you gave some background, we can answer > better). > > > On 30/07/13 12:44, Rose Beck wrote: > >> Just like there was a paper for Jena SDB ("Efficient RDF Storage and >> Retrieval in Jena2"), is there any paper for Jena TDB? >> > > "Efficient RDF Storage and Retrieval in Jena2" is what we might call > "RDB", and is no longer part of the Jena project. SDB is a more efficient > (and different mapping) to SQL. It exists for enviroments where using SQL > is required. TDB scales better and executes faster. > > > On Tue, Jul 30, 2013 at 5:06 PM, Chris Dollin >> <[email protected]>**wrote: >> >> On Tuesday, July 30, 2013 04:41:09 PM Rose Beck wrote: >>> >>>> Does Jena TDB support reification. If yes, then how? >>>> >>> > Note that named graphs enable many of the uses of reification because a > graph can be treated as quoted. As statements often don't come in units of > one, named graph provide a way to handle groups of statements. > > What are you trying to achieve? > > > What do you mean by "support reification"? >>> >>> Also can you please give an example SPARQL query plan illustrating the >>>> >>> use >>> >>>> of reification in Jena. >>>> >>> >>> There's nothing in SPARQL specific to reification. You just use the RDF >>> reification vocabulary. >>> >> > And indeed the experience with the design was that there is a major > tradeoff of handling reification specially and performance. Just look at > the query in 5.1. > > SPARQL is more complicated than RDQL. Proper handling can restrict the > amount of work sent to the database engine, requiring multiple JDBC > operations per SPARQL query. > > That design loaded at maybe 1000 triples/s for the first 1m or so triples > : TDB loads at 30K for 100m, with bursts of 75+K in the first million. TDB > is same-process. > > Andy > > > >>> Chris >>> >>> -- >>> "It does not need to take events in their correct order." >>> /Hexwood/ >>> >>> Epimorphics Ltd, http://www.epimorphics.com >>> Registered address: Court Lodge, 105 High Street, Portishead, Bristol >>> BS20 >>> 6PT >>> Epimorphics Ltd. is a limited company registered in England (number >>> 7016688) >>> >>> >>> >> >
