When the Data Integration Service runs a mapping, it generates queries against the mapping sources and performs operations on the query results based on the mapping logic and the transformations within the mapping. The queries and operations often include predicate expressions. Predicate expressions represent the conditions that the data must satisfy. The filter and join conditions in Filter and Joiner transformations are examples of predicate expressions.

With the predicate optimization method, the Data Integration Service also attempts to apply predicate expressions as early as possible in the mapping to improve mapping performance.

The Data Integration Service infers relationships from by existing predicate expressions and creates new predicate expressions. For example, a mapping contains a Joiner transformation with the join condition "A=B" and a Filter transformation with the filter condition "A>5." The Data Integration Service might be able to add "B>5" to the join condition.

The Data Integration Service applies the predicate optimization method with the early selection optimization method when it can apply both methods to a mapping. For example, when the Data Integration Service creates new filter conditions through the predicate optimization method, it also attempts to move them upstream in the mapping through the early selection method. Applying both optimization methods improves mapping performance when compared to applying either method alone.

The Data Integration Service applies the predicate optimization method if the application increases performance. The Data Integration Service does not apply this method if the application changes the mapping results or it decreases the mapping performance. The Data Integration Service applies this optimization method by default.