Multiple Query Execution through SQL Rewriting

Current database systems typically process queries in a query-at-a-time fashion without considering common subexpressions that can be shared. Such a design leads to suboptimal processing, i.e., lower throughput, higher latency, and poor scalability. Taking advantage of sharing common subexpressions in batches or streams of concurrent queries is one way to mitigate those problems. Another option is to implement new shared operators in database engines, or rewrite them completely. The main aspect of our work is that we achieve shared query execution without modifying any database engines. In this work we present a novel approach that completely relies on SQL query rewriting and already existing features of today’s database engines by minimizing redundant work. We first describe three different methods how to exploit sharing opportunities. Secondly, the methods are evaluated via microbenchmarks on four different common database systems. Our results show that it is very well possible to achieve a significant reduction of execution cost without modifying a database engine. In several cases our methods improve the performance by orders of magnitude