A data backup method, network node and system

Abstract

An embodiment of the invention discloses a data backup method, network nodes and a system which are used for optimizing synchronous backup schemes of write buffers. According to the data backup method, the network nodes and the system, persistent management synchronization mechanisms are added, and write buffer structures of slave nodes are optimized. The method comprises the steps of writing incremental data sent by master nodes into write buffers of the slave nodes according to a preset time sorting order; writing incremental data files sent by the master nodes into disks of the slave nodes, wherein the incremental data files are generated by the master nodes according to persistent data and disk data files of the master nodes, and the persistent data come from write buffers of the master nodes and are to be written into disks of the master nodes; according to the incremental data files, deleting data related to the incremental data files in the write buffers of the slave nodes. The data backup method, the network nodes and the system are implemented, so that hardware expenses of the slave nodes can be saved, and then hardware costs for establishment of data service clusters can be reduced.

Description

technical field [0001] The invention relates to the technical field of data storage, in particular to a data backup method, network node and system. Background technique [0002] In this field, in order to improve data service capabilities and resist data loss risks, multiple data service nodes are used to form a data service cluster to provide data services to clients. The data service cluster backs up data to multiple nodes. When a node fails, other nodes can provide the same data service as the failed node in time and assist in recovering the failed node. Therefore, the data service cluster has a higher disaster recovery capability than a single node. [0003] At present, the data service cluster mainly adopts the write buffer synchronous backup scheme based on the master / slave (Master / Slave) architecture. In this solution, one node is selected as the master node to undertake the main data service, and other nodes are used as slave nodes to assist the master node in pro...

AI Technical Summary

Problems solved by technology

[0013] 1. In order to maintain the organizational structure based on the content sorting order in the write buffer, when incremental data is written into the write buffer of the slave node, it needs to be compared and merged with the original data, and repeated operations increase the computing overhead of the processor ;

[0014] 2. When the slave node triggers to write the data in the write buffer to the disk, a large amount of processor calculation and temporary memory overhead will be generated;

[0015] 3. When data is written from the disk of the node, multiple scattered storage areas are generated. It is necessary to organize the multiple scattered storage areas into a continuous storage area by means of background disk data sorting, and delete redundant data files. This processing generates a lot of disk IO (In / Out, input and output), processor calculation and memory overhead

Images