W3C home > Mailing lists > Public > semantic-web@w3.org > January 2006

Re: tips for estimating the size of an sql database in rdf

From: Steve Harris <S.W.Harris@ecs.soton.ac.uk>
Date: Tue, 10 Jan 2006 11:32:06 +0000
To: Semantic Web <semantic-web@w3.org>
Message-ID: <20060110113206.GA18941@login.ecs.soton.ac.uk>

On Tue, Jan 10, 2006 at 12:16:38 +0100, Henry Story wrote:
> I am interested if anyone has good tips to help estimate how large a  
> plain old legacy sql database is going to be in triples for a  
> particular ontology. I can probably make them up, but who knows,  
> perhaps someone has already written a calculator for this. This would  
> be very useful in estimating what projects should would be initial  
> candidates for putting online behind a SPARQL interface, given that  
> current specialised triple databases can take between 20 million and  
> 300 million triples. [1]

It's not something I've much experience with, but, if the database is in
3rd normal form I would expect the typical equivalent RDF representation
to be (C-1) * R triples per table, where C is num. of columns, and R is
num. of rows. that assumes that the primary key column is subsumed as the
subject of the triples, and the other columns are represented as objects.
But it would surely vary a lot with the database schema.

- Steve
Received on Tuesday, 10 January 2006 11:32:15 UTC

This archive was generated by hypermail 2.4.0 : Tuesday, 5 July 2022 08:44:55 UTC