Aaron Elmore: Adventures in Database Compression

University of Chicago

Abstract: Columnar databases enable effective compression by improving entropy through attribute locality and provides opportunities for fast query execution directly on compressed data. In this talk I will briefly overview how compressed query execution works in columnar systems and discuss techniques developed by our group over the past several years. This includes a pattern-inferred attribute decomposition for improved string compression and query performance, a bounded float compression technique for fast filtering on limited precision numeric data, and partially ordered dictionary compression.