Doctoral Thesis: Instance-Optimized Database Indexes and Storage Layouts
Modern database systems are faced with the challenge of maintaining high performance while keeping pace with both increasingly diverse use cases and increasingly large data volumes. Current methods for handling this challenge are unsatisfactory—one method is to custom-build a database system for each new application, but these bespoke systems require years of intense engineering effort and are only achievable by large corporations with significant resources. An alternative method is to tune the configuration knobs of a pre-built, general-purpose database system to fit the new application, but the performance impact of tuning is limited by the design of the pre-built database system, and even a well-tuned general-purpose database cannot reach the performance of a custom-built database.
In this thesis, we explore a fundamentally different way of maintaining high performance for diverse use cases: instance-optimized database systems are built to be able to automatically self-adjust in order to achieve the best performance for a given use case. We introduce novel instance-optimized versions of two fundamental database components—indexes and data storage layouts—that outperform existing state-of-the-art indexes and data layouts by orders of magnitude. We also demonstrate how to incorporate multiple instance-optimized database components into an end-to-end analytic database system that outperforms a well-tuned commercial cloud-based analytics system by up to 3X.
- Date: Monday, August 1
- Time: 2:00 pm - 3:00 pm
- Location: 32-D463
Additional Location Details:
Thesis Supervisor: Prof. Tim Kraska