MPP database management method, device and system

Abstract

The embodiment of the present application provides an MPP database management method, device and system. In the process of backing up each backup object, the snapshot of each record information in each backup object page corresponding to the backup object is recorded as a snapshot backup Identifies that only the modified record information is recorded in the snapshot backup space allocated for the backup target page, and then the modification status of each backup target page is recorded through the first bitmap space, and the backup target page is recorded in the second bitmap space The modification status of each record information. In this way, during the backup process, only the modified record information, snapshot ID, and some modification status need to be recorded, and there is no need to backup the full amount of data files, which can effectively save the storage space consumed during backup and reduce the time consumed by backup. It is possible to reduce the impact on the upper layer business. In addition, a separate backup can be implemented for data objects without affecting the normal business operations of other data objects.

Description

technical field [0001] The present application relates to the technical field of databases, in particular, to an MPP database management method, device and system. Background technique [0002] At present, in the data backup and recovery scheme for MPP (Massive Parallel Processing, large-scale parallel processing) database, the PITR (Point-In-Time-Recovery, failure recovery based on point in time) backup and recovery scheme is usually adopted. In the backup and recovery scheme, a basic backup is first required to back up the full data files. When the data volume of the data files is large, each full-level backup will consume a huge amount of storage space. [0003] In addition, basic backup in the case of a large amount of data will also consume a lot of time in the entire backup process, and although the backup recovery scheme can be restored to any point in time between the start of the backup and the end of the backup, it also requires Reapplying archived log information...

AI Technical Summary

Problems solved by technology

[0002] At present, in the data backup and recovery scheme for MPP (Massive Parallel Processing, large-scale parallel processing) database, the PITR (Point-In-Time-Recovery, failure recovery based on point in time) backup and recovery scheme is usually adopted. In the backup and recovery plan, a basic backup is first required to back up the full data files. When the data volume of the data files is large, each full-level backup will consume a huge amount of storage space.

[0003] In addition, basic backup in the case of a large amount of data will also consume a lot of time in the entire backup process, and although the backup recovery scheme can be restored to any point in time between the start of the backup and the end of the backup, it also requires Reapplying archived log information is also very time-consuming in the process of reapplying

[0004] At the same time, in this backup and recovery scheme, only the overall backup and recovery of the database can be performed each time, and the backup and recovery of a specific object (such as a large table) cannot be realized, resulting in less flexibility in the entire backup or recovery process. Difference

Images