Towards instance-optimized data systems

In recent years, we have seen increased interest in applying machine learning to system problems. For example, there has been work on improve query optimization, indexing, storage layouts, scheduling, log-structured merge trees, sorting, compression, and sketches, among many other data management tasks. Arguably, the ideas behind these techniques are similar: is used model and/or workload order derive a more efficient algorithm or structure. Ultimately, will allow us build "instance-optimized" systems: that is, systems self-adjust given distribution provide unprecedented performance without need for tuning by an administrator. While of promise orders-of-magnitude better lab settings, still general skepticism about how practical current really are. The following intended as progress report ML Systems its readiness real-world deployments, with focus our projects done part Data AI Lab (DSAIL) at MIT By no means it comprehensive overview all existing work, which steadily growing over past several years not only database community but also systems, networking, theory, PL, adjacent communities.