No DBA? No Regret! Multi-Armed Bandits for Index Tuning of Analytical and HTAP Workloads With Provable Guarantees

Automating physical database design has remained a long-term interest in research due to substantial performance gains afforded by optimised structures. Despite significant progress, majority of today's commercial solutions are highly manual, requiring offline invocation administrators (DBAs). This status quo is untenable: identifying representative static workloads no longer realistic; and tools remain susceptible the query optimiser's cost misestimates. Furthermore, modern application environments like hybrid transactional analytical processing (HTAP) systems render modelling next impossible. We propose self-driving approach online index selection that does not depend on DBA optimiser, instead learns benefits viable structures through strategic exploration direct observation. view problem as one sequential decision making under uncertainty, specifically within bandit learning setting. Multi-armed bandits balance exploitation provably guarantee average converges policies optimal with perfect hindsight. Our comprehensive empirical evaluation against state-of-the-art tuning tool demonstrates up 75% speed-up 59% HTAP environments. Lastly, our framework outperforms Monte Carlo tree search (MCTS)-based providing 24% speed-up.