Data query method and cluster database system

Abstract

The invention discloses a cluster database system. The cluster database system comprises a query analyzer and a plurality of database nodes. The query analyzer is used for analyzing a received query request, resolving the received query request into a query task stream and generating an execution rule; and meanwhile, the query analyzer is used for distributing one or more than one query task in the query task stream for corresponding database nodes according to the execution rule. The database nodes are used for executing corresponding query tasks together with other relevant database nodes according to the execution rule after the distributed query tasks are received; and besides, the database nodes are used for combining query data obtained through execution of a certain task and query data obtained through execution of the same query task by the other relevant database nodes so as to obtain query result data, and the data is the final query result data corresponding to the query request or the data is used as input data for the next query task and therefore the plurality of database nodes can carry out the query task stream according to the execution rule to obtain the final query result data conveniently.

Description

technical field [0001] The present invention relates to the technical field of data processing, and more specifically, to a data query method, a cluster database system and a database node. Background technique [0002] Message Passing Interface (MPI) is a programming model for message passing, a standard and specification for parallel programming, and is widely used in high-performance computing and parallel computing. MPI has high communication performance, good program portability and powerful functions. At present, the implementation of MPI mainly includes MPICH (jointly developed by Argonne National Laboratory and Mississippi State University), OpenMP (proposed by the OpenMP Architecture Review Board) and LAM (Local Area Multicomputer, developed by Ohio State University for heterogeneous computing. MPI package), etc. Using MPI to run parallel programs can facilitate inter-process communication and synchronous control without having to care about the details of communi...

AI Technical Summary

Problems solved by technology

[0005] (1) If the amount of data is large enough, the extracted data cannot be stored in a single data node

Therefore, this method of extracting first and then querying and calculating is not suitable for application scenarios with a huge amount of data

[0006] (2) For global statistics and sorting queries, the two-stage query processing method is adopted, and the parallel computing capabilities between multiple nodes cannot be fully utilized when the results are merged and calculated in the second stage

[0007] (3) The two-stage query processing method cannot solve complex query processing, such as the associated query of large tables distributed on multiple database nodes and the nested query of large tables

Images