Intelligent monitoring and load scheduling method based on deep learning

Abstract

The invention discloses an intelligent monitoring and load scheduling method based on deep learning. The method comprises the following steps: acquiring information of all in-use servers in real time,summarizing the information to a central server, analyzing the information through a deep learning model by the central server, and evaluating the conditions of the in-use servers so as to estimate the weight of each server, sending the weight of the in-use server estimated by the central service in real time to the reverse proxy server, wherein each client requests; and enabling the reverse proxy service to calculate the server address of the optimal solution according to the current weight information; and enabling the reverse proxy server to establish a service communication channel between the client and the server of the address according to the optimal server address. According to the invention, the load of the server is evaluated more accurately in combination with deep learning,so that the reverse proxy loads the request of the client to each server more accurately and intelligently, and the bottleneck of the traditional load balancing mode is solved.

Description

technical field [0001] The invention relates to the technical field of artificial intelligence, in particular to an intelligent monitoring and load scheduling method based on deep learning. Background technique [0002] Currently, load balancing implementations on the market simply allocate client requests to different servers based on traffic or business complexity. Common allocation methods include round-robin allocation, ordinary hash algorithm allocation, consistent hash algorithm allocation, etc. Among the above load distribution methods, only the consistent hash algorithm distribution method takes into account the situation of server downtime, but it is not enough to only consider the situation of server downtime. Because the server downtime in the actual production environment is relatively rare compared to the fact that the actual performance of the server itself is different due to the difference in hardware configuration and software operation and cannot really ac...

AI Technical Summary

Problems solved by technology

Among the above load distribution methods, only the consistent hash algorithm distribution method takes into account the situation of server downtime, but it is not enough to only consider the situation of server downtime

Because the server downtime in the actual production environment is relatively rare compared to the fact that the actual performance of the server itself is different due to the difference in hardware configuration and software operation, and it cannot really achieve the purpose of load.

The actual situation is that some server hardware configurations have stronger performance and more idle resources than other servers, but they handle the same concurrency pressure as other servers, and cannot give full play to their performance advantages. The performance is poor, and the server with less remaining resources handles a large number of concurrency at full capacity

In turn, some server resources are tight and some server resources are idle, so that the actual load balance (load imbalance) cannot be achieved.

Images