Analysis of the current state of functioning of server clusters of information systems and load management methods

Abstract

The article analyzes the current state of functioning of the server infrastructure of information systems and methods of managing the load on server clusters. The main factors affecting the efficiency of work are identified: the growth in the number and complexity of client requests, limited hardware resources, the impossibility of scaling the infrastructure, as well as the insufficient performance of existing load balancing methods. It was found that most modern algorithms have high computational complexity, low resistance to failures and significant dependence on the quality of training data. The study considered classical and intellectual approaches to the optimal use of cluster system resources. Adaptive methods based on game theory, in particular Nash equilibrium, were recognized as the most effective. This approach allows you to formalize the interaction of servers with requests without the need for training, ensuring resistance to deviations. The scientific novelty lies in using Nash equilibrium for optimal load distribution taking into account the current state of servers and the possibility of task decomposition. The proposed method is the basis for creating new adaptive load balancing systems.