Analytics-Driven Lossless Data Compression for Rapid In-situ Indexing, Storing, and Querying

The analysis of scientific simulations is highly data-intensive and is becoming an increasingly important challenge. Peta-scale data sets require the use of light-weight query-driven analysis methods, as opposed to heavy-weight schemes that optimize for speed at the expense of size. This paper is an attempt in the direction of query processing over losslessly compressed scientific data. We propose a co-designed double-precision compression and indexing methodology for range queries by performing unique-value-based binning on the most significant bytes of double precision data (sign, exponent, and most significant mantissa bits), and inverting the resulting metadata to produce an inverted index over a reduced data representation. Without the inverted index, our method matches or improves compression ratios over both general-purpose and floating-point compression utilities. The inverted index is light-weight, and the overall storage requirement for both reduced column and index is less than 135%, whereas existing DBMS technologies can require 200-400%. As a proof-of-concept, we evaluate univariate range queries that additionally return column values, a critical component of data analytics, against state-of-the-art bitmap indexing technology, showing multi-fold query performance improvements.