Research On The Multi-model Query Processing

Big data management aims to establish data hubs that support data in multiple models and types in an all-around way.Thus,the multi-model database system is a promising architecture for building such a multi-model data store.The integrated data center faces many problems in multi-model query processing.For example,the existing query language semantics are not rich enough,the existing index tuning technologies are poor to deal with complex multi-model scenarios,and there is no mature multi-model query optimization algorithm.In order to improve the efficiency of multi-model query processing,this paper studies the key technologies of multi-model query processing,including the design of a unified and flexible multi-model query language Multi-SQL,the research on general automatic index tuning technologies for complex scenarios,and the efficient multi-model query optimization algorithm.So far,Multi-SQL is the first unified query language designed based on a multimodal perspective.It realizes the unified management of multi-model data and considers the collaborative processing of multi-model data.It is an extensible and practical query language.Can be easily extended to adapt to more complex scenarios.First of all,this article gives the usage of Multi-SQL through some specific cases.Then,this paper gives the concrete formal definition of multi-modal definition language and multi-modal manipulation language,and analyzes the implementation framework of Multi-SQL,and theoretically explains its semantic diversity.In addition,this paper also studies the general index self-tuning technology based on deep learning and deep reinforcement learning(CNNIS and DRIS),which can fully model the database.Compared with manual or rule-based methods,the method in this paper can be considered More comprehensive and more timely.The two general index recommendation methods proposed in this paper use CNN and deep reinforcement learning to solve the problem of automatic index recommendation,starting from the original data and from the perspective of database feature modeling.Finally,this paper studies a two-level multi-modal query optimization algorithm based on automatic mining of dependencies(Two-Optimal),which can make full use of the existing query optimization capabilities of the underlying engine,and can greatly improve query execution efficiency.Moreover,this article has been tested on both public test benchmarks and self-generated test benchmarks.Compared with the existing query engine,Two-Optimal has increased by 7.34%.