Load balancing method, device, node and computer-readable storage medium

Abstract

Embodiments of the present invention provide a load balancing method, device, node, and computer-readable storage medium, which relate to the field of distributed storage. The method and device determine the basic weight of each storage space based on the storage space remaining rate of each storage node ; Judging whether the selection probability information of the storage node satisfies the weight correction condition, if so, modify the basic weight of the storage node according to the correction parameters to obtain the target weight of the storage node; if not, use the basic weight of the storage node as The target weight of the storage node; according to the target weight of all the storage nodes, a priority interval sequence of the storage node is formed; the target storage node with the highest priority in the priority interval sequence of the storage node is obtained as the storage node for data writing Compared with the prior art, the present invention can avoid load unbalance phenomenon of the distributed system caused by improper selection of storage nodes during the data writing process.

Description

Load balancing method, apparatus, node, and computer-readable storage medium technical field The present invention relates to the field of distributed storage, in particular, to a load balancing method, device, node and computing device computer readable storage medium. Background technique [0002] With the rapid growth of Internet access and data traffic, the workload of distributed storage systems is gradually increasing. increase, resulting in greater processing pressure on the nodes in the system. When data is written to the distributed storage system, due to different sections The service capabilities of the nodes are different, so the tasks assigned to each node are quite different, which will lead to storage The capacity difference is large, the data distribution in the cluster is uneven, and the load of the cluster and nodes is not balanced. In order to solve the load imbalance phenomenon, the prior art uses disk utilization, central processing unit CPU (Ce...

AI Technical Summary

Problems solved by technology

When data is written into the distributed storage system, due to the different service capabilities of different nodes, the tasks assigned to each node are quite different, which will lead to a large difference in the storage capacity of each node, and the data distribution in the cluster is not consistent. Uniform, the cluster and nodes have unbalanced load

[0003] In order to solve the unbalanced load phenomenon, in the prior art, performance indicators such as disk utilization, central processing unit (Central Processing Unit, CPU) occupancy and memory occupancy are combined as the weight of the node, and the node corresponding to the maximum weight is selected. The node is used as the node for writing data. When performing dynamic load balancing, this method does not aim at disk space utilization. It only selects available nodes based on the level of some performance indicators, which may cause the node to be duplicated. Being selected makes the disk full in advance, and other nodes have no chance to be selected, so that in the process of data writing, the load of the distributed system cannot reach a balanced state due to improper selection of storage nodes

Images