Sunday, October 14, 2012

Status update

A small status update.

The SPARQL 1.0 implementation is around 80% done, missing still support for some of the corner cases in nested OPTIONALs with references to upper level variables. The nice part is that Josef is almost done with all the core SPARQL functions, leaving little more to be done. Timea is also preparing the Sesame provider layer for Quest that will let us offer access to Quest through Sesame's console to setup a proper SPARQL end-point.

Also, even though we haven't fully finished the implementation it is already in a good state to cover most queries, so Mindaugas is setting up the environment for running the Berlin Benchmark. Expect some number comparing MySQL D2RQ, Virtuoso virtual RDF graphs and Quest (running on top of MySQL) very very soon.

Talking about benchmarks, the paper by Bail et al. [1] was accepted. They compare MySQL, Virtuoso virtual graphs and Quest in virtual RDF mode. Some great numbers for Quest, up to 500 times faster than Virtuoso, almost as fast as native MySQL. It seems that Virtuoso virtual graphs performance drops dramatically with each join you add to the queries, 4 queries already set Virtuoso to 20 or less queries per second (qps) when Quest and MySQL are on the hundreds (500-900). Five joins and we go down to 1 qps with virtuoso, with 550 qps for Quest and 840 qps for MySQL. The paper is available here [2].

Exciting times!

[1] Samantha Bail, Sandra Alkiviadous, Bijan Parsia, David Workman, Mark van Harmelen, Rafael S. Gonçalves, and Cristina Garilao. FishMark: A Linked Data Application Benchmark. SSWS, 2012. ISWC 2012 workshop. To appear.

No comments:

Post a Comment