Construction of a simulation model for managing load on computing nodes in a server cluster based on the theory of fuzzy logic and Nash equilibrium

This study’s object is the process of managing (balancing) the load on the computing nodes in a server cluster of information systems. The task addressed is to enable optimal (even) distribution of server resources within a cluster system. A simulation model of the process that distributes the load on computing nodes of a server cluster has been built, based on an improved model of the optimal use of server resources in a cluster based on Nash equilibrium and an improved method of adaptive load balancing in cluster systems according to Nash equilibrium. The simulation model has been constructed on the basis of fuzzy logic theory to determine the feasibility of decomposing tasks into subtasks, as well as game theory, in particular Nash equilibrium, to determine the optimal distribution of tasks/subtasks across cluster system servers for their parallel processing. The model built makes it possible to improve the efficiency (uniformity) of server resource distribution by an average of 13% compared to the classical load management method based on Nash equilibrium, by 61% with the Round Robin method, and by 63% with the Least Connection method throughout the entire process of cluster operation. This, in turn, allows for a more than 2-fold increase in the number of processed tasks from client requests compared to the above load balancing methods. In addition, the impact of the load balancing process on the processing time of tasks/subtasks by the cluster system servers was estimated. Based on the results of simulation modeling, it can be concluded that the application of the devised model does not exceed the total allowable processing time (no more than 315 ms) of tasks/subtasks from client requests compared to existing load balancing methods