Distribution keys and collocated joins

If you have tables that are usually joined, distribute them on the same column that you use to join the tables. The IBM® Netezza® system can then perform a collocated join that minimizes data movement and provides optimal performance.

The Netezza architecture distributes processing across many individual SPUs each with its own dedicated memory and data slices. These individual processors operate in a “shared nothing” environment that eliminates the contention for shared resources which occurs in a traditional SMP architecture. In a collocated join, each SPU can operate independently of the others without network traffic or communication between SPUs.