通信学报 ›› 2021, Vol. 42 ›› Issue (12): 1-16.doi: 10.11959/j.issn.1000-436x.2021229

• 学术论文 •    下一篇

QML:一种混合空间索引结构

崔栋, 温巧燕, 张华, 王华伟   

  1. 北京邮电大学网络与交换技术国家重点实验室,北京 100876
  • 修回日期:2021-11-30 出版日期:2022-12-25 发布日期:2021-12-01
  • 作者简介:崔栋(1991- ),男,山东潍坊人,北京邮电大学博士生,主要研究方向为学习索引、移动互联网安全
    温巧燕(1959- ),女,陕西户县人,博士,北京邮电大学教授、博士生导师,主要研究方向为现代密码理论、量子密码与量子信息、网络安全技术等
    张华(1978- ),女,吉林四平人,博士,北京邮电大学副教授、博士生导师,主要研究方向为网络安全、隐私保护等
    王华伟(1989- ),男,河南驻马店人,博士,北京邮电大学在站博士后,主要研究方向为区块链安全、去中心化的身份安全认证技术
  • 基金资助:
    国家自然科学基金资助项目(62072051);国家自然科学基金资助项目(61976024);国家自然科学基金资助项目(61972048);中央高校基本科研业务费专项资金资助项目(2019XD-A01);中华人民共和国教育部区块链重点项目计划基金资助项目(2020KJ010802)

QML: a hybrid spatial index structure

Dong CUI, Qiaoyan WEN, Hua ZHANG, Huawei WANG   

  1. The State Key Laboratory of Networking and Switching Technology, Beijing University of Posts and Telecommunications, Beijing 100876, China
  • Revised:2021-11-30 Online:2022-12-25 Published:2021-12-01
  • Supported by:
    The National Natural Science Foundation of China(62072051);The National Natural Science Foundation of China(61976024);The National Natural Science Foundation of China(61972048);The Fundamen-tal Research Funds for the Central Universities(2019XD-A01);The Key Project Plan of Blockchain in Ministry of Education of the People’s Republic of China(2020KJ010802)

摘要:

为了丰富现有学习多维索引的功能并提高索引效率,提出了可以保留数据分布特征的动态数据分段算法DDSA,并结合四叉树和Z顺序曲线构建了混合空间索引(QML),在此基础上分别设计范围查询算法和KNN查询算法。这种保留数据分布特征的索引可以灵活实现快速查询和更新。实验结果表明,QML 索引在实现丰富功能的前提下优化了检索效率,数据更新的时间复杂度为O(1)。与R*-tree相比,QML索引存储减少约33%,更新效率提升40%~80%。查询效率与最优树形索引相近。

关键词: 数据库, 空间索引, 学习索引

Abstract:

In order to enrich the functionalities of existing learned multidimensional indexes and improve the efficiency, the dynamic data segmentation algorithm DDSA was proposed, which could preserve the data distribution characteristics.A hybrid spatial index was constructed by combining the QuadTree and Z-order curve (QML).The range query algorithm were designed and KNN query algorithm respectively.The proposed index allowed flexible fast queries and updates with preserving the characteristics of data distribution.Experimental results show that QML optimizes the query efficiency on the premise of achieving rich functionalities, and the time complexity of data update is O(1).Compared with R*-tree, the storage consumption of QML is reduced by about 33%, and the update efficiency is improved by 40%~80% .The query efficiency is similar to the optimal tree Index.

Key words: database, spatial index, learned index

中图分类号: 

No Suggested Reading articles found!