Learning from SQL: Database Agnostic Workload Management

Abstract

Database Management Systems largely ignore the wealth of information present in SQL query workloads. In this work, we present a vision for database agnostic workload management. We start by providing an architecture for the SQLShare platform, a database-as-a-service built for researchers with minimal database experience. We demonstrate how we used this system to collect and publish a diverse workload of hand-written SQL queries to aid database research in general and workload analytics in particular. We also provide an analysis of the SQLShare workload and using the learnings from this analysis, we present the design of Querc, a database-agnostic workload management and analytics service, describe potential applications, and show that separating workload representation from labeling tasks affords new capabilities and can outperform existing solutions for representative tasks, including workload sampling for index recommendation, user labeling for security audits, error prediction for debugging, and query runtime prediction for resource allocation.