Scalability of Oracle RDF repository

Number of Triples that can be stored: The limit is only the database size (in Oracle 10g Release 2 data sizes that can be stored is in the exabyte range). The bytes required to store a triple in the database is approximately equivalent to the size in bytes of the triple itself. In short, billions of triples can be stored.

Query performance: Query performance scales well and is in the order of seconds for millions of triples (on a machine with Red Hat Enterprise Linux AS 3 System with one 3.06 GHz Pentium 4 CPU and 2048 MB of main memory). Some figures for triple sizes upto 80M are at http://www.oracle.com/technology/tech/semantic_technologies/pdf/vldb_2005.pdf (can also be accessed by going to http://www.oracle.com/technology/tech/semantic_technologies/index.html and clicking on 'An Efficient SQL-based RDF querying scheme', the second link under 'What's new').

Please note that tuning the database helps improve performance, and also an upcoming patch release has significant performance improvements compared to 10.2.0.2.

Load performance: On a machine with Red Hat Enterprise Linux AS 3 System with one 3.06 GHz Pentium 4 CPU and 2048 MB of main memory the load performance is about 35 minutes to load a million triples. We are planning significant improvements of this in an upcoming database release.

We are planning on creating a technical white paper with updated performance numbers for upcoming releases. I will post that once it is available.

Melli

PS: Apologies for the delayed response, which was due to the fact that I was on vacation.

OWL support is planned for our next major release. Details will be made available later this Fall. Stay tuned.

Melli