Method and device for processing database redo data

Abstract

Embodiments of the present invention provide a method and an apparatus for processing redo data in a database, where the method includes: generating, by each application thread of multiple application threads included in a server, redo data according to a database modification operation, accordingly saving the redo data in a buffer allocated to each application thread, saving an identifier of the application thread in a time sequence queue after a time sequence queue lock is acquired, and releasing the time sequence queue lock after the saving ends; and determining, by a data reading thread, that a data reading condition is satisfied, reading a sequence of the identifiers of the application threads from the time sequence queue, successively reading a piece of redo data from the buffer of each application thread corresponding to the identifier of each application thread in the time sequence queue, and writing the piece of redo data to a redo queue. With use of the present invention, redo data processing efficiency can be improved by separating a time sequence queue from a data queue, thereby increasing a concurrent throughput of a database system.

Description

technical field [0001] The invention relates to the technical field of communication network, in particular to a method and a device for processing database redo data. Background technique [0002] The redo log (redolog) is an important guarantee for data reliability in database operations, and can provide more data security protection than the file system. Its basic principle is that when the database modification operation is submitted, the data will not be written into the data file on the disk immediately, but stored in the memory, so as to avoid writing into the data file every time the database update request is required, thereby improving the performance of the database. Update performance; and generate redo data (redodata) according to the modification operation and write it to the disk to become a redo log, so as to avoid data loss due to the loss of memory data in the event of a failure. When the database operation process fails, the data can be recovered through ...

AI Technical Summary

Problems solved by technology

Because redo data has timing constraints, when multiple threads copy redo data to the redo queue, the timing needs to be guaranteed within the protection scope of the global redo queue lock, that is, redo must be preempted every time redo data is written. Time-consuming data copy after queue lock

[0004] In the existing technology, multiple threads can only generate and copy redo data in a serial manner, so there is a defect of low redo data processing efficiency, which affects the throughput of the system under concurrent application of large business volume

Images