A data query method

Abstract

The invention discloses a data inquiring method. When data tables are stored in a distributed mode, an application server horizontally splits two related tables according to the same subtable number, subtables, with the same index number, of the two tables are hashed to a distributed database server according to the same hash formula, and the database server can carry out simple data inquiring and processing according to the correlation of the subtables during data inquiring. By means of the method, network transmission cost of cross-machine query can be avoided during union query, processing time delay of complex query is reduced, and the method is suitable for OLTP application scenes.

Description

technical field [0001] The invention relates to the data processing technology of the distributed database system, in particular to a data query method in the distributed database system. Background technique [0002] At present, with the scale of data and its access volume, relational databases pose great challenges: billions of records, several terabytes of data, thousands of TPS, and tens of thousands of QPS make traditional relational databases overwhelmed. Simple hardware upgrades It has been impossible to solve the problem. However, the current distributed database technology cannot effectively support complex relational queries. [0003] To solve this problem, it is usually used to first split the data table horizontally, and then distribute the split sub-tables to multiple servers to realize large-scale data storage. Specifically: select one or several fields in the data table as the split key, and hash the records into different database servers. When the client ...

AI Technical Summary

Problems solved by technology

[0004] The above method can realize the distributed storage of data tables with a large record scale. However, cross-machine queries are prone to occur when performing table association queries, that is, sub-tables that need to be associated are distributed on different servers. In this case, on the one hand, each database The server cannot perform merged query processing, and the application server needs to read the sub-tables from each database server before performing the corresponding merged query processing, which will lead to a large data processing load on the application server. The more data there is, the worse the performance of the system will be; on the other hand, since the data in the database server needs to be sent to the application server for processing, correspondingly, redundant network data transmission will also be caused.

[0005] Therefore, under the distributed database system, the traditional data query method has problems such as large network transmission overhead and prolonged processing time, and cannot meet the associated query in the online transaction processing (OLTP) application scenario with high latency requirements.

A method suitable for relational queries in the OLTP context has not yet been proposed

Images