This document proposes improvements to the query performance of triple stores that store RDF data in a relational database. It explores storing RDF triples in different orders in three tables and developing a query rewriting scheme. It also looks at optimizing the physical schema through graph clustering techniques that aim to reduce the number of disk accesses by grouping related triples closer together. The paper presents an implementation of these techniques over a million triples and shows they can yield significant performance benefits on complex queries.