Analytical Review of the Methods of Dynamic Load Balancing under Conditions of Uncertainty in the Execution Time of Tasks

computationally complex problems can be solved using distributed computing resources, but the problem arises in the optimal way of distributing tasks between computing nodes in order to reduce the total execution time. However, the computation time for a task on the same device is not constant and can change dynamically over time. In this article, we explore methods of dynamic load balancing in order to minimize the computation time of the problem. We study three groups of methods based on the use of probability theory and mathematical statistics methods, evolutionary algorithms and artificial neural networks. We have shown that methods based on artificial neural networks can reduce the computation time of the problem and minimize the complexity of the dynamic load balancing algorithm.