Self-adaptive load balancing method based on host practical loads and load balancing device and system

Abstract

The invention discloses a self-adaptive load balancing method based on host practical loads. The method is performed in a load balancing device. The load balancing device is in communication connection with a plurality of server hosts and stores prediction models for the number of sessions capable of being processed by servers under different performance indexes. The method comprises the steps of collecting current performance indexes of the servers and reconstructing the current performance indexes into index feature vectors; inputting the index feature vectors of the servers into the corresponding prediction models, thereby obtaining the number of the sessions capable of being processed by the servers under the current performance index; and setting load sharing weights for the servers according to the number of the sessions capable of being processed by the servers and carrying out corresponding flow distribution according to the load sharing weights of the servers. The invention also discloses a corresponding load balancing device and system.

Description

technical field [0001] The invention relates to the technical fields of computers and the Internet, in particular to an adaptive load balancing method, device and system based on the actual load of a host. Background technique [0002] Load balancing (also known as load sharing) is a set of servers composed of multiple servers in a symmetrical manner. Each server has an equivalent status and can provide external services independently without the assistance of other servers. Through some kind of load sharing technology, the requests sent from the outside are distributed to multiple servers or network devices, such as Web servers, FTP servers, enterprise key application servers and other mission-critical servers, etc., and the server that receives the request responds to the client independently request, thereby improving the business processing capability and ensuring high availability of the business. [0003] The commonly used load balancing methods of database cluster sy...

AI Technical Summary

Problems solved by technology

The polling method selects back-end nodes in turn to execute user query requests. Although it can make each back-end have tasks, it cannot achieve the optimal allocation of each back-end task

The weighted round-robin method assigns different weights to different database nodes, but it usually uses a static weight method to distribute traffic to the target server in a fixed proportion. It needs to rely on manual experience to specify load sharing parameters, scalability, flexibility, and accuracy. lower sex

Images