Domain vector accelerator (DVA): A query accelerator for relational operations

A fast method is described for performing relational operations and, in particular, for an equijoin between two relations that stand in a one-to-many relationship. The method is based on a bit-vector technique called domain vector acceleration (DVA). The approach to join acceleration is described and compared analytically with two other join accelerators, hybrid-hash join and join indices. Results show that using domain vectors for simple, binary equijoins between very large tables significantly improves the efficiency of the equijoin operation. Domain vectors also reduce the amount of data that must be cached on disk, relative to join indices and materialized views. DVA can be applied to a wide variety of relational operations, including select, interest, union, semijoin and outer join