Back to index

Join Processing in Database Systems with Large Main Memories

Leonard D. Shapiro

Summary by: Steve Gribble and Armando Fox

One-line summary: Four join algorithms are presented and analyzed (sort-merge, simple hashing, GRACE hashing, hybrid hashing), and hybrid hashing is shown to dominate the others in performance in most situations, assuming a large enough memory (order of square root of smaller relation).

Overview/Main Points


Convincingly argues for the use of hash-based join processing in database systems. Considering that physical memory is even more in abundance than when this paper was written, the arguments have probably gotten stronger over time, although I wonder how much data set size (measured in number of tuples) has grown relative to physical memory size growth.


Back to index