When the Integration Service processes a Joiner transformation, it reads rows from both sources concurrently and builds the index and data cache based on the master rows. The Integration Service then performs the join based on the detail source data and the cache data. The number of rows the Integration Service stores in the cache depends on the partition type, the source data, and whether you configure the Joiner transformation for sorted input. To improve performance for an unsorted Joiner transformation, use the source with fewer rows as the master source. To improve performance for a sorted Joiner transformation, use the source with fewer duplicate key values as the master.