Memory cluster storage balancing method and device

Abstract

The invention discloses a memory cluster storage balancing method and device, and relates to the field of database. The method comprises the steps that a surface partition segment of an original first level key K1 is segmented to obtain patch numbers; according to the original level key K1, and the patch numbers, a balanced level key K1' is generated; the (K1', DO) data storage form is adopted to store the key value data; the original first level key K1 is the original logic key for the customized object DO, the original level key K1 comprises a database segment, a data table segment, and a table partition section, the balanced first level key K1' is a balanced logic key for the customized object DO. Thus the uniform distribution of the data in a memory cluster data node is achieved, and the memory capacity of the memory cluster storage is fully exploited.

Description

technical field [0001] The invention relates to the field of databases, in particular to a memory cluster storage balancing method and device. Background technique [0002] The proxy server performs a consistent hash on the keys in the cluster, ideally ensuring that the keys are evenly distributed on the cluster nodes. Because the data volume of partitions is usually different, the different data volumes of partitions will lead to unbalanced storage resource occupancy of cluster nodes. [0003] Such as figure 1 As shown, assume that the cluster has 4 nodes, and the storage of each node is 10. Store data in 8 partitions twice, store 4 partitions (4 keys) for the first time, and the data volume is 1, 2, 4, 8 respectively; store 4 partitions (4 keys) for the second time, the data volume Both are 2. If the partition data stored twice are evenly stored on 4 nodes, then the data volumes of the 4 nodes are 3, 4, 6, and 10 respectively. If you insert data into the cluster nodes...

AI Technical Summary

Problems solved by technology

[0004] It can be seen that the traditional consistent hash algorithm does not consider the impact of partitions, and directly processes data members on an equal basis. At the same time, the way of adding virtual nodes in recent years has exacerbated the uneven distribution of partitions. Therefore, it is necessary to solve the problem of excessive concentration of partition data on several services. Nodes cause overall performance bottlenecks

Images