2023 KR QueryBooster: Improving SQL Performance Using Middleware Services for Human-Centered Query Rewriting

[DelaramRajaei]

Reason to read this paper

I selected this paper primarily because of its focus on query rewriting. Although its main goal is in enhancing SQL performance, it offers valuable insights into query rewriting and serves as a useful guide for crafting well-written papers on the subject.

---

Main problem

In database systems, developers are tasked with using their own understanding and knowledge to rephrase the provided queries from the application before sending them to the database. However, the existing solutions do not offer the flexibility required for query rewriting according to human needs and preferences.

This paper focuses on developing an easy-to-use middleware between application and database, to rewrite queries for better performance.

---

Related Works &Their Gaps

• Native rewriting plugins: Users are not given the option to modify or rewrite queries.
• Third-party solutions: Rewrite queries by adding hints to help the database optimizer generate faster plans based on their estimations and search methods: (None of them let users create their own rewriting rules.)
• Commercial systems: Helps with query rewriting for applications on top of databases (None of them let users create their own rewriting rules)
• General pattern-matching tools: Not designed for SQL queries and can be used for rewriting codes as well. They do not consider the unique semantics.

---

Proposed Method

QueryBooster: A middleware interface for users to formulate rewriting rules.
First, there's an easy-to-use rule language where users can define the query pattern and its replacement. They can add extra conditions and actions to make the rules more complex. Second, users can just show examples of the original query and how they want it rewritten.

There is no need to change the application or the database. They only need to replace the existing DB connector.

---

Input/Output

Input: SQL Query
Output: Rewritten Query based on the user's rules and examples.

---