Although new data management systems scale more efficiently, for example, by decoupling storage and compute, they still necessitate significant performance engineering. As such, several researchers have proposed “learned” database components, including learned indexes, query optimizers, etc. However, many of these tools focus on a single problem (e.g., designing index structures) and/or only provide recommendations, which a DBA or data engineer must implement. In contrast, Pavlo et al. propose a “self-driving” DBMS that can “configure, manage, and optimize itself automatically” based on a specific objective function (i.e., that defines throughput, latency, and availability requireme…