Automatic optimization of array queries

Non-trivial scientific applications often involve complex computations on large multi-dimensional datasets. Using relational database technology for these datasets is cumbersome since expressing the computations in terms of relational queries is difficult and time-consuming. Moreover, query optimization strategies successful in classical relational domains may not suffice when applied to the multi-dimensional array domain. The RAM (Relational Array Mapping) system hides these issues by providing a transparent mapping between the scientific problem specification and the underlying database system. This paper focuses on the RAM query optimizer which is specifically tuned to exploit the characteristics of the array paradigm. We detail how an intermediate array-algebra and several equivalence rules are used to create efficient query plans and how, with minor extensions, the optimizer can automatically parallelize array operations