Recent years have witnessed a constant growth in the amount of RDF data available on the Web. This growth is largely based on the increasing rate of data publication on the Web by different actors such as governments, life science researchers or geographical institutes. RDF data generation is mainly done by converting already existing legacy data resources into RDF (e.g. converting data stored in relational databases into RDF), but also by creating that RDF data directly (e.g. sensors). These RDF data are normally exposed by means of Linked Data-enabled URIs and SPARQL endpoints.

Given the sustained growth that we are experiencing in the number of SPARQL endpoints available, the need to be able to send federated SPARQL queries across them has also grown. Tools for accessing sets of RDF data repositories are starting to appear, differing between them on the way in which they allow users to access these data (allowing users to specify directly what RDF data set they want to query, or making this process transparent to them). To overcome this heterogeneity in federated query processing solutions, the W3C SPARQL working group is defining a federation extension for SPARQL 1.1, which allows combining, in a single query, graph patterns that can be evaluated in several endpoints.

In this PhD thesis, we describe the syntax of that SPARQL extension for providing access to distributed RDF data sets and formalise its semantics. We adapt existing techniques for distributed data access in relational databases in order to deal with SPARQL endpoints, which we have implemented in our federation query evaluation system (SPARQL-DQP). We describe the static optimisation techniques that we implemented in our system and we carry out a series of experiments that show that our optimisations significantly speed up the query evaluation process in presence of large query results and optional operators.