Towards Generating HiFi Databases

Anupam Sanghi, Rajkumar Santhanam, Jayant R. Haritsa

Abstract: Generating synthetic databases that capture essential data characteristics of client databases is a common requirement for database vendors. We recently proposed Hydra, a workload-aware and scale-free data regenerator that provides statistical fidelity on the volumetric similarity metric. A limitation, however, is that it suffers poor accuracy on unseen queries. In this paper, we present HF-Hydra (HiFi-Hydra), which extends Hydra to provide better support to unseen queries through (a) careful choices among the candidate synthetic databases and (b) incorporation of metadata constraints. Our experimental study validates the improved fidelity and efficiency of HF-Hydra.