Efficient Main-Memory Algorithms for Set Containment Join Using Inverted Lists

We present two algorithms for set containment joins based on inverted lists. The first algorithm scans the left relation and determines for each tuple all the qualifying tuples by querying the inverted file for the right relation. The second algorithm employs the common inverted file for both relations. We focus on improving performance of algorithms in main memory by reducing number of L2 cache misses which is achieved by applying such techniques as partitioning and compression. We study algorithms analytically and experimentally and determine which one is better depending on parameters of the input relations. We also demonstrate that both algorithms are superior to some other known methods for set containment joins.