Davide Lanti: Data Scaling for OBDA Benchmarking

When: Friday, March 11, at 14:30
Where: Turing building, room Gilles Kahn
Abstract:
Ontology-based data access (OBDA) is a novel approach whose aim is to facilitate the access to legacy databases to the end-users through the help of a mediating ontology.

Although some prototype implementations for OBDA exist, performance is still an open issue for their widespread adoption.

In order to test the performance of OBDA in real-world applications, suitable benchmarks are needed. These benchmarks should come with an ontology, queries over terms in the vocabulary of the ontology, a database instance, and mappings that relate each
term in the vocabulary of the ontology to a query over the database. These dimensions together form an input for the OBDA system. Additionally, a benchmark should allow for a scalability analysis with respect to one or more of these dimensions.

In this talk we shall focus on how to provide instances for a scalability analysis with respect to the database dimension. In particular, we explore the problem of scaling an initial database instance so as to obtain scaled instances of different sizes that display some properties identified over the initial database instance. These properties should be relevant to the task of performance testing with respect to the given ontology, mappings, and queries. Moreover, the properties have to be simple enough so that efficient generation of large instances displaying the properties is still possible.

Comments are closed