The main fine tuning strategies used by relational database administrators are the construction of access structures, such as indexes, partial indexes and materialized views, and techniques such as denormalization and query rewriting. These techniques and access structures, together or separately, can improve the performance of queries submitted to the database. Database partitioning, a technique traditionally used for data distribution, has also the potential for fine tuning, since it allows the scanning of tables to be performed only on partitions that satisfy query predicates. Even in queries with high selectivity predicates, whose execution plans often use indexes, partitioning can offer even greater benefit. This dissertation proposes to evaluate the partitioning as a fine tuning action of relational databases and, for that, develops heuristics for selection of partitioning strategies and evaluation of its benefit. An evaluation of the quality of the results obtained is carried out through experiments with a standard benchmark for this type of research and we have shown that, in certain cases, it is advantageous to partition data.