The web is changing the way in which data warehouses are designed, used, and queried. With the advent of initiatives such as Open Data and Open Government, organizations want to share their multidimensional data cubes and make them available to be queried online. The RDF data cube vocabulary (QB), the W3C standard to publish statistical data in RDF, presents several limitations to fully support the multidimensional model. The QB4OLAP vocabulary extends QB to overcome these limitations, allowing to im- plement the typical OLAP operations, such as rollup, slice, dice, and drill-across using standard SPARQL queries. In this paper we introduce a formal data model where the main object is the data cube, and define OLAP operations using this model, independent of the underlying representation of the cube. We show then that a cube expressed using our model can be represented using the QB4OLAP vocabulary, and finally we provide a SPARQL implementation of OLAP operations over data cubes in QB4OLAP.