Schedule - PostgreSQL Development Conference 2024

Advice is seldom welcome but efficacious

Date: 2024-05-29
Time: 10:30–11:20
Room: Canfor(1600)
Level: Intermediate

Advisory features that provide users with advice and suggestions have been attracting attention recently, as these features make it easier for people to make decisions and perform tasks that previously required extensive experience and know-how.

One well-known advisory feature is the Index Advisor, which suggests indices that may improve query performance. However, there are also a variety of other good advisory features in PostgreSQL, such as those related to configuration parameters, partitioning, and execution plans. For example, the execution plan advisor (pg_plan_advsr) I developed previously suggests improvements for inefficient execution plans generated by complex queries with many joins. It also tells you how to write hint clauses, since it utilizes hint clauses (pg_hint_plan) to explore a plan space internally. If users know and can master these tools, they can gain significant benefits, such as improved query performance.

In this talk, I will introduce the advisory features available in PostgreSQL worldwide, and then focus on the execution plan advisor and explain the new function“Suggestion feature for Extended Statistics”with performance verification results. And finally, aiming to advance PostgreSQL, I will provide feedback to the PostgreSQL development community on findings from the new feature development and performance verification results.

This presentation will cover the following topics:

1. List of advisory features available in PostgreSQL (extensions, etc.)

2. Introduction of new features of the execution plan advisor
    - Overview of the tool (benefits, mechanism, examples of use)
    - The new“Suggestion feature for Extended Statistics”and the results of performance verification
    - Future roadmap

3. Feedback to the development community for the advancement of PostgreSQL
    - Expanding where extended statistics apply

The following people will benefit from this talk:

Those who are interested in 
    - advisory features available in PostgreSQL
    - developing advisory features in the future
    - utilizing extended statistics


Tatsuro Yamada