Adventures With Triplestores
by Ash Smith
Since its inception, the Southampton Open Data service has been run on 4Store. Reasons are varied, but it certainly helped that several of its key developers used to work at Southampton, and we have their mobile numbers! However, we recently began a project to see if we can replace it with something a bit more up-to-date, as our poor little server is starting to choke under increased load. Semantic Web technology has evolved, and we wanted to make sure we stay at the cutting edge.
As an extra incentive, none of us really had any experience setting up a triplestore, and we’ve had several requests for RDF-based databases from students, all of which we have had to decline. The University’s central IT department and well as Electronics and Computer Science both have procedures in place for students or researchers who need web space or relational databases for their projects, but it’s a shortcoming of our service that we can’t currently offer triplestores for those who need them. So we decided to start installing some triplestores on some Ubuntu machines and see what happens.
The experiments are still ongoing, but our first candidate was the excellent Sesame. This introduced an interesting problem – none of us are big fans of Java, nor have any of us used Apache Tomcat before. Tomcat is a Java-based web server and its ‘servlet container’ component is a requirement of Sesame, as well as many other triplestore back-ends. Being a team consisting of two system programmers and one ex-researcher who never really got the hang of Java, we were immediately overwhelmed by phrases like “deploy the WARs” and the comprehensive, if verbose, detail of Sesame’s official documentation.
Help was at hand thanks to this article by Blogger user Nadinima. Once we’d realised that installing Tomcat 7 on Ubuntu was a simple case of apt-get install tomcat, this blog post made the rest quite straightforward, and I got Sesame installed and configured in under an hour, despite my lack of Tomcat knowhow.
The next phase of the tests has been to modify my publisher tool Hedgehog so that it supports Sesame as well as 4Store. A positive side-effect of experimenting with different server architectures is that Hedgehog will become more versatile as a result. Currently Hedgehog is happily running on our development server, importing data into 4Store and Sesame simultaneously. We still need to do some stress-testing and then compare the performance to other stores, such as Parliament and Longwell, but I’ll modify Hedgehog appropriately if we discover it doesn’t work with them. Once we pick a favourite, we’ve got to work out how we’re going to transition to the new server architecture, which will probably involve lots of messing about with VMs and the writing of a custom SPARQL wrapper. But hopefully in doing so we can educate ourselves on the installation and configuration of triplestores, and also build a server image for those who want a triplestore set up quickly with a view to offering this as a standard service to members of the University.