Graph data is central to many applications, ranging from social networks to scientific databases. Graph formats maximize the flexibility offered to data designers, as they are mostly schema-less and thus can be used to capture very heterogeneous-structure content. RDF, the W3C's format for sharing open (linked) data, adds the possibility to attach semantics to data, describing application-domain constraints by means of ontologies; in turn, this leads to implicit data that is also part of a graph even if it is not explicitly in it. In this paper, we present a structured walk through the problem of analyzing and exploring RDF graphs by finding groups of structurally similar nodes, and by automatically identifying interesting aggregates theirein. We outline the challenges raised by such processing in large, complex RDF graphs, outline the basic principles behind existing solutions, and highlight opportunities for future research.