IA2 Preprocessing: Establishing the Foundation for Index Selection

Table of Links

Abstract and 1. Introduction

1. Related Works

   2.1 Traditional Index Selection Approaches

   2.2 RL-based Index Selection Approaches

2. Index Selection Problem

3. Methodology

   4.1 Formulation of the DRL Problem

   4.2 Instance-Aware Deep Reinforcement Learning for Efficient Index Selection

4. System Framework of IA2

   5.1 Preprocessing Phase

   5.2 RL Training and Application Phase

5. Experiments

   6.1 Experimental Setting

   6.2 Experimental Results

   6.3 End-to-End Performance Comparison

   6.4 Key Insights

6. Conclusion and Future Work, and References

5.1 Preprocessing Phase

The preprocessing phase is critical for establishing a solid foundation for IA2’s operation. It consists of two components:

\ Workload Model: Enhanced by the underlying optimizer and what-if cost models, the workload model captures database workload variabilities. It integrates four essential components: Query Plan features, reflecting database reactions; current index configurations; Meta information about database configurations and budget; and embedded tokenized queries. This model is crucial for providing accurate state representations to the downstream DRL training task, significantly boosting IA2’s generalization capabilities across diverse workloads.

\ Index Candidates Enumerator: This component extends beyond exhaustive enumeration, employing validation rules and restrictions to discern the relevance among queries. By leveraging permutations and heuristic rules that cater to generic operators and workload structures, the Enumerator crafts index candidates. This approach, inspired by and integrating advancements from Lan et al. [7] , enriches the selection pool with a broader array of indexing strategies, poised to optimize performance across varying scenarios. The generated index candidates form the raw action space for the downstream RL task, laying a foundational step for IA2’s decision-making process in selecting optimal indexes

\

:::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.

:::

\