Adaptive Action Pruning: Scaling Index Selection for Unseen Workloads | HackerNoon

Table of Links

Abstract and 1. Introduction

2. Related Works

   2.1 Traditional Index Selection Approaches

   2.2 RL-based Index Selection Approaches

3. Index Selection Problem

4. Methodology

   4.1 Formulation of the DRL Problem

   4.2 Instance-Aware Deep Reinforcement Learning for Efficient Index Selection

5. System Framework of IA2

   5.1 Preprocessing Phase

   5.2 RL Training and Application Phase

6. Experiments

   6.1 Experimental Setting

   6.2 Experimental Results

   6.3 End-to-End Performance Comparison

   6.4 Key Insights

7. Conclusion and Future Work, and References

6.4 Key Insights

Summarizing our extensive experiments, IA2 represents a significant advancement in index selection, outperforming existing methods in several key areas:

Rapid Training Efficiency: IA2 excels with its unparalleled training speed, leveraging a what-if cost model and pre-trained models to facilitate quick adaptability and learning. This efficiency allows IA2 to drastically reduce training time compared to competitors, making it highly suitable for environments where speed is crucial.

Advanced Workload Modeling: Unlike static or exhaustive methods, IA2 employs dynamic workload modeling, enabling it to adapt to changing database queries and structures seamlessly. This flexibility ensures optimal index selection across diverse scenarios, including previously unseen workloads.

Effective Action Space Exploration: IA2 introduces an innovative approach to pruning and navigating the action space, efficiently identifying meaningful actions early in the training process. This strategy contrasts with the more resource-intensive techniques of SWIRL [6] or the rigid rules of Lan et al. [7], offering a balanced pathway to optimizing index configurations without exhaustive search or oversimplification.

:::info
Authors:

(1) Taiyi Wang, University of Cambridge, Cambridge, United Kingdom ([email protected]);

(2) Eiko Yoneki, University of Cambridge, Cambridge, United Kingdom ([email protected]).

:::

:::info
This paper is available on arxiv under CC BY-NC-SA 4.0 Deed (Attribution-Noncommercial-Sharelike 4.0 International) license.

:::