With the development of the Semantic Web, an increasingly large number of organizations represent their data in RDF format. A single machine cannot efficiently process complex queries on RDF graphs. It becomes necessary to use a distributed cluster to store and process large-scale RDF datasets that are required to be partitioned. In this paper, we propose a semantic-aware partitioning method for RDF graphs. Inspired by the PageRank algorithm, classes in the RDF schema graphs are ranked. A novel partitioning algorithm is proposed, which leverages the semantic information of RDF and reduces crossing edges between different fragments. The extensive experiments on both synthetic and real-world datasets show that our semantic-aware RDF graph partitioning outperforms the state-of-the-art methods by a large margin.