Date of Award:
Doctor of Philosophy (PhD)
Kevin R. Moon
A database is used to store and retrieve data, which is a critical component for any software application. Databases requires configuration for efficiency, however, there are tens of configuration parameters. It is a challenging task to manually configure a database. Furthermore, a database must be reconfigured on a regular basis to keep up with newer data and workload. The goal of this thesis is to use the query workload history to autonomously configure the database and improve its performance. We achieve proposed work in four stages: (i) we develop an index recommender using deep reinforcement learning for a standalone database. We evaluated the effectiveness of our algorithm by comparing with several state-of-the-art approaches, (ii) we build a real-time index recommender that can, in real-time, dynamically create and remove indexes for better performance in response to sudden changes in the query workload, (iii) we develop a database advisor. Our advisor framework will be able to learn latent patterns from a workload. It is able to enhance a query, recommend interesting queries, and summarize a workload, (iv) we developed LinkSocial, a fast, scalable, and accurate framework to gain deeper insights from heterogeneous data.
Sharma, Vishal, "Deep Learning Data and Indexes in a Database" (2021). All Graduate Theses and Dissertations. 8214.
Copyright for this work is retained by the student. If you have any questions regarding the inclusion of this work in the Digital Commons, please email us at .