RDF query optimization is a challenging problem. Although considerable factors and their impacts on query efficiency have been investigated, this problem still needs further investigation. We identify that decomposing query into a series of light-weight operations is also effective in boosting query processing. Considering the linked nature of RDF data, the correlations among operations should be carefully handled. In this paper, we present SGDQ, a novel framework that features a partition-based Summary Graph Driven Query for efficient query processing. Basically, SGDQ partitions data and models partitions as a summary graph. A query is decomposed into subqueries that can be answered without inter-partition processing. The final results are derived by perform summary graph matching and join the results generated by all matched subqueries. In essence, SGDQ combines the merits of graph match processing and relational join-based query implementation. It intentionally avoids maintain huge intermediate results by organizing sub-query processing in a summary graph driven fashion. Our extensive evaluations show that SGDQ is an effective framework for efficient RDF query processing. Its query performance consistently outperforms the representative state-of-the-art systems.