METHODS OF OPTIMIZING THE DISTRIBUTION OF THE LOAD ON THE COMPUTING RESOURCE OF THE CLOUD SERVICE INFRASTRUCTURE

Abstract

Methods and algorithms for optimizing the distribution of the load on the computing resource of the cloud service infrastructure are investigated in the paper. It is noted that load balancing is a major challenge among cloud networks. The main purpose of load balancing is to use resources efficiently and improve performance. Along with this, it removes nodes that contain heavy load as well as nodes that are not working properly or performing a small task. It is emphasized that the following can be identified as basic criteria related to increasing the efficiency of cloud load balancing in real time: minimization of resource movement costs and task execution costs, maximization of transfer speed and task execution. The quality (efficiency) of balancing in the work is understood as an integral criterion that includes the essential parameters of the system's operation. It is emphasized that the mathematical model of dynamic distribution of virtual resources to physical machines in cloud computing, which provides accounting of the previous state of the system load and the effect of the appearance of a new resource on the load balance in the system, and is distinguished by the use of a load adjustment factor to achieve balancing. It is noted that the genetic algorithm for the optimal distribution of new virtual resources is distinguished by the implementation of a tree-like structure of chromosomes with the preservation of highly loaded nodes, which ensures an increase in the quality of load balancing and a reduction in the dynamic movement of resources. It is emphasized that the multi-criteria optimization mathematical model of task scheduling in cloud computing ensures the minimization of task transfer time, execution time and execution costs, which differs by considering the parameters of the channel between the user and the data center.