Overview
Database management systems (DBMS) are software systems that control the creation, maintenance, and use of databases, i.e., organized collections of data. Relational DBMS (RDBMS) are incredibly ubiquitous today -- they underlie technology used by most people every day if not every hour. As a consequence, the database system course is widely offered in major universities around the world as part of the undergraduate computer science degree program. A core component of this course is the design and implementation of the query optimizer module, which is responsible for efficient execution of an SQL query. It is fundamentally a very complex component involving non-trivial disk-based algorithms and search space exploration techniques. Consequently, it is traditionally considered as a difficult component to fathom by an adult learner taking an undergraduate-level database course. Given a SQL query, a student would typically like to understand how it is executed on the underlying RDBMS by studying the associated query execution plan (QEP) generated by the RDBMS. However, every commercial database vendor has its own secret sauce for the implementation of the query optimizer. Consequently, understanding a QEP and how a query optimizer generated such a plan demand not only deep knowledge of various query processing-related concepts but also vendor-specific implementation details. We advocate that this can be challenging for learners learning database systems for the first time. The TRUSS project aims to addresses this problem by developing learner-friendly tools and techniques to facilitate learning of various facets of relational query processing. We advocate that these tools should supplement the traditional modes of learning (e.g., lectures, textbooks). The project’s goals are not only technological in nature; there is a pervasive desire to stimulate a cultural shift in the way that database systems education is undertaken globally.
This research is partially funded by NTU Tier 1 grant.
Key Achievements
- ARENA and MOCHA (ACM SIGMOD 2023, VLDB 2022) are the world's first systems that allow a learner to explore interesting alternative query plans for his/her query.
- LANTERN (ACM SIGMOD 2021, 2022, ACM SIGCSE TS 2025) is the world's first neural-rule-based hybrid system that allows a learner to interact with the QEP of an SQL query using natural language.
- NEURON (ACM SIGMOD 2019) is the world's first rule-based system that allows a learner to interact with the QEP of an SQL query using natural language.