基于异步Dueling DQN和计划时间预测网络的连接优化器

高瑞玮; 乔少杰; 韩楠; 闵圣捷; 李贺; 覃晓; 张桃; 元昌安

doi:10.12263/DZXB.20220490

您当前的位置：

首页 >

文章列表页 >

基于异步Dueling DQN和计划时间预测网络的连接优化器

学术论文 | 更新时间：2025-12-08

- 基于异步Dueling DQN和计划时间预测网络的连接优化器
- A Join Optimizer Based on Asynchronous Dueling DQN and Plan Latency Prediction Network
- 电子学报 2023年51卷第7期页码：1868-1874
- 作者机构：
  
  1.成都信息工程大学软件工程学院,四川成都 610225
  2.成都信息工程大学管理学院,四川成都 610103
  3.中电智元数据科技有限公司,北京 100081
  4.西安电子科技大学计算机科学与技术学院,陕西西安 710126
  5.南宁师范大学,广西南宁 530001
  6.宜宾学院,四川宜宾 644000
  7.广西科学院,广西南宁 530007
- 作者简介：
  
  [ "高瑞玮　男，1996年12月出生于四川射洪，现为成都信息工程大学软件工程学院硕士研究生.主要研究方向为人工智能数据库.E-mail: 1013384678@qq.com" ]
  [ "乔少杰(通讯作者)　男，1981年10月出生于山东招远，博士，现为成都信息工程大学软件工程学院教授.主要研究方向为人工智能数据库、时空数据库、机器学习.E-mail: sjqiao@cuit.edu.cn" ]
- 基金信息：
  
  国家自然科学基金(62272066;61962006);四川省科技计划资助(2021JDJQ0021;2022YFG0186);教育部人文社会科学研究规划基金(22YJAZH088);成都市“揭榜挂帅”科技项目(2022-JB00-00002-GX;2021-JB00-00025-GX);成都市技术创新研发项目(2021-YF05-02413-GX;2021-YF05-02414-GX);中国电子科技集团公司第五十四研究所高校合作课题(SKX212010057);成都信息工程大学科技创新能力提升计划(KYTD202222)
- DOI：10.12263/DZXB.20220490
  中图分类号： TP311
- 收稿：2022-05-05，
  
  修回：2023-02-18，
  
  纸质出版：2023-07-25
- 稿件说明：
移动端阅览
高瑞玮,乔少杰,韩楠等.基于异步Dueling DQN和计划时间预测网络的连接优化器[J].电子学报,2023,51(07):1868-1874.

GAO Rui-wei,QIAO Shao-jie,HAN Nan,et al.A Join Optimizer Based on Asynchronous Dueling DQN and Plan Latency Prediction Network[J].ACTA ELECTRONICA SINICA,2023,51(07):1868-1874.
高瑞玮,乔少杰,韩楠等.基于异步Dueling DQN和计划时间预测网络的连接优化器[J].电子学报,2023,51(07):1868-1874. DOI： 10.12263/DZXB.20220490.

GAO Rui-wei,QIAO Shao-jie,HAN Nan,et al.A Join Optimizer Based on Asynchronous Dueling DQN and Plan Latency Prediction Network[J].ACTA ELECTRONICA SINICA,2023,51(07):1868-1874. DOI： 10.12263/DZXB.20220490.

摘要

连接顺序选择是查询优化领域中极具挑战性的研究方向，对于数据库管理系统获得良好的查询性能至关重要.然而，传统优化方法和现有智能优化方法均存在着不足，如规划时间过长、容易得到质量较差的连接计划、编码未考虑结构特征、依赖基数估计和代价估计使得连接计划无法反映真实的执行时间等.针对上述问题，提出了一种新型基于异步Dueling DQN（Deep Q-network）和计划时间预测网络的连接优化器：ADP-Join（Asynchronous Dueling DQN and Plan Latency Prediction Network for Join Order Selection）.ADP-Join集成了一种新的编码方法，能够区分不同结构的连接计划.ADP-Join设计了计划时间预测网络PLN（Plan Latency Prediction Network）来改善现有基于强化学习优化器的奖励机制.再者，提出异步更新机制改进Dueling DQN模型来提升训练性能和减少训练时间.大量的实验结果表明，在TPC-H和JOB真实数据集上ADP-Join的性能优于现有的智能优化器.

Abstract

Join order selection is a challenging research topic in the field of query optimization

and it is very important for database management system to obtain good query performance. However

both traditional optimization methods and existing intelligent optimization methods have disadvantages such as long planning time

easily to obtain poor quality join plans

encoding without considering structural characteristics

making join plans unable to reflect the real execution time due to dependency on cardinality estimation and cost estimation. In order to solve the above problems

a new join optimizer ADP-Join (Asynchronous Dueling DQN and Plan latency prediction network for Join order selection) is proposed. ADP-Join integrates a new encoding method that can distinguish join plans of different structures. ADP-Join designs a plan latency prediction network to improve the reward mechanism of existing reinforcement learning-based optimizers.Furthermore

the asynchronous update mechanism is proposed to improve the Dueling DQN model to improve the training performance and reduce the training time. Extensive experimental results show that ADP-Join outperforms existing intelligent optimizers on real TPC-H and JOB datasets.

关键词

Keywords

references

QIAO S J , YANG G P , HAN N , et al . Cardinality estimator: Processing SQL with a vertical scanning convolutional neural network [J]. Journal of Computer Science and Technology , 2021 , 36 ( 4 ): 762 - 777 .

LI G L , ZHOU H X , GAO L . AI meets database: AI4DB and DB4AI [C]// Proceedings of SIGMOD Conference . New York : ACM , 2021 : 2859 - 2866 .

LI G L , ZHOU X H , SUN J . openGauss: Anautonomous database system [J]. Proceedings of VLDB Endowment , 2021 , 14 ( 12 ): 3028 - 3041 .

MARCUS R , PAPAEMMANOUIL O . Towards a hands-free query optimizer through deep learning [EB/OL]. ( 2018 )[2022]. https://api.semanticscholar.org/CorpusID:52880119 https://api.semanticscholar.org/CorpusID:52880119 .

YU X , LI G L , CHAI C L . Reinforcement learning with tree-LSTM for join order selection [C]// Proceedings of 36th International Conference on Data Engineering . Washington : IEEE , 2020 : 1297 - 1308 .

ZHOU H X , LI G L , CHAI C L . A learned query rewrite system using Monte Carlo tree search [J]. Proceedings of VLDB Endowment , 2021 , 15 ( 1 ): 46 - 58 .

KAOUDI Z . ML-based cross-platform query optimization [C]// Proceedings of the 36th International Conference on Data Engineering . Washington : IEEE , 2020 : 1489 - 1500 .

MARCUS R . Deep reinforcement learning for join order enumeration [C]// Proceedings of the First International Workshop on Exploiting Artificial Intelligence Techniques for Data Management . New York : ACM , 2018 : 31 - 34 .

KRISHNAN S , YANG Z H , GOLDBERG K , et al . Learning to optimize join queries with deep reinforcement learning [EB/OL].( 2018 )[2022]. https://api.semanticscholar.org/CorpusID:51952349 https://api.semanticscholar.org/CorpusID:51952349 .

WANG J Y , CHAI C L , LIU J B . FACE: A normalizing flow based cardinality estimator [J]. Proceedings of VLDB Endowment , 2021 , 15 ( 1 ): 72 - 84 .

MARCUS R , NEGI P , MAO H Z . Neo: A learned query optimizer [J]. Proceedings of VLDB Endowment , 2019 , 12 ( 11 ): 1705 - 1718 .

MNIH V , KAVUKCUOGLU K , SILVER D . Human-level control through deep reinforcement learning [J]. Nature , 2015 , 518 ( 7540 ): 529 - 533 .

DEY R , SALEM F M . Gated recurrent unit neural networks gate-variants [C]// Proceedings of 60th International Midwest Symposium on Circuits and Systems . Washington : IEEE , 2017 : 1597 - 1600 .

POSS M , FLOYD C . New TPC benchmarks for decision support and web commerce [J]. SIGMOD Record , 2000 , 29 ( 4 ): 64 - 71 .

LEIS V , GUBICHEV A , MIRCHEV A . How good are query optimizers, really? [J]. Proceedings of VLDB Endowment , 2015 , 9 ( 3 ): 204 - 215 .

ZHANG J , ABEDJAN Z , HOSE K . AlphaJoin: Join order selection à la AlphaGo [C]// Proceedings of the VLDB 2020 PhD Workshop Co-located with the 46th International Conference on Very Large Databases . Trondheim : VLDB Endowment , 2020 : 5 - 8 .

WASS F . Join order selection [C]// Proceedings of British National Conference on Databases . Berlin : Springer , 2000 : 256 - 265 .

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

基于图组合优化的高效社区搜索

知识数据协同的多对手智能空中博弈策略设计

基于强化学习的免调参即插即用单光子图像重建方法

基于强化学习的离散事件系统最优定向监控

基于强化学习的自免疫动态攻击生成方法