In-Memory Deep Vectorization

Oracle Database optimizes hash joins by pushing operations down into the scans of both sides of the join.

In deep vectorization, the left scan performs the build, while the right scan conducts the probe and gathers and projects rows back to the parent of the hash join operator. Joins can account for a large percentage of SQL execution time for data warehouse workloads. Improving performance of hash joins by 100% or more can significantly improve performance.

Related Topics