Despite the increasing popularity of RDF as a data representation method, there is no accepted measure of the importance of nodes in an RDF graph. Such a measure could be used to sort the nodes returned by a SPARQL query or to find the important concepts in an RDF graph. In this paper we propose a graph-theoretic measure called noc-order for ranking nodes in RDF graphs based on the notion of centrality. We illustrate that this method is able to capture interesting global properties of the underlying RDF graph using study cases from different knowledge domains. We also show how well noc-order behaves even if the underlying data has some noise, i.e. superfluous and/or erroneous data. Finally, we discuss how information about the importance of different predicates either based on their informativeness, prior semantic information about them or user preferences can be incorporated into this measure. We show the effects of such modifications to the ranking method by examples.