Introduction

This work try to provide a tool for DBAs to visilaize and see why query plans generated by learned query optimizers outperform those generated by traditional optimizers.

Design targets:

1. multiple exploration modes (query-centric vs. rulecentric)
2. support for multiple performance metrics (latency, number of page spills)
3. visual comparison of query plans
4. support for interactive retrieve the relevant information.

System

It uses exposed knobs to steer query optimization.

It measures the query based on wall time, the number of page accesses, and the memory footprint.

---

• ← Previous Post
• Next Post →