Maliva: Using Machine Learning to Rewrite Visualization Queries Under Time Constraints

We consider data-visualization systems where data is stored in a database, and a middleware layer translates a frontend request to a SQL query to the database to compute visual results. We focus on the problem of handling visualization requests with predetermined time constraints. We study how to rewrite the original query by adding hints and/or conducting approximations so that the total time is within the time constraint. We develop a novel middleware solution called Maliva, which adopts machine learning (ML) techniques to solve the problem. It applies the Markov Decision Process (MDP) model to decide how to rewrite queries and uses training instances to learn an agent that can make a sequence of decisions judiciously for an online request. Our experiments on both real and synthetic datasets show that compared to the baseline approach that relies on the original SQL query, Maliva performs significantly better in terms of both the chance of serving requests interactively and query execution time.