The scale of RDF graph grows very rapidly. Managing huge scale RDF graph distributively is becoming increasingly important. Partitioning RDF graph is a vital pre-processing step for the goal. When applying graph partitioning algorithms developed over past decades to RDF graph represented using well known RDF model such as Directed Labeled Graphs, Bipartite Graph, the vertices which a triple depends on may be in different partitions. Such partitioning on the RDF models induces huge communication overhead during processing queries. We argue in this paper that there is need for a representation of RDF to enable the parallel and distributed computing application on RDF data. We propose statement hyper graph model which avoid this crucial deficiency of the graph model of RDF data. The proposed models reduce the decomposition problem to the well-known hyper graph partitioning problem. In the light of this model, we explore the cases like horizontal partitioning, vertical partitioning, grid partitioning, etc and evaluate their performance.