Efficient Execution of Multi-Query Data Analysis Batches Using Compiler Optimization Strategies

Henrique Andrade, Suresh Aryangat, Tahsin Kurc, Joel Saltz, Alan Sussman

To appear at 16th Workshop on Languages and Compilers for Parallel Computing (LCPC03), College Station, TX, 2-4 October 2003

Full Text, Printable Abstract.

Abstract

This work investigates the leverage that can be obtained from compiler optimization techniques for efficient execution of multi-query workloads in data analysis applications. Our approach is to address multi-query optimization at the algorithmic level, by transforming a declarative specification of scientific data analysis queries into a high-level imperative program that can be made more efficient by applying compiler optimization techniques. These techniques -- including loop fusion, common subexpression elimination and dead code elimination -- are employed to allow data and computation reuse across queries. We describe a preliminary experimental analysis on a real remote sensing application that analyzes very large quantities of satellite data. The results show our techniques achieve sizable reductions in the amount of computation and I/O necessary for executing query batches and in average execution times for the individual queries in a given batch.