Dynamic Algorithm for Load Distribution Between Computing Resources of Cloud Environments

Abstract

The article considers the development of a dynamic algorithm for distributing and balancing the load, characterized by nonlinear parameters and fractal structure, between the computing resources of cloud data centers (DCs). A rationale is given for the application of methods of nonlinear dynamics and chaos theory to solving this problem, which is based on the topological similarity of the behavior of deterministic chaotic systems and the input network load of the data center. The load randomness test was conducted by calculating Lyapunov indices. To eliminate noise components of network traffic, the principal component method of its time series was applied. The problem of phase space reconstruction is solved by the delay method based on the embedding theorem and the average mutual information method. Based on this reconstruction, a method for predicting the load dynamics was developed. It was presented in the form of a system of discrete mappings connected by an algebraic polynomial, providing successful prediction of irregular time series and forming the basis for constructing a matrix for distributing data center resources. The aim of the article is to increase the efficiency of using the resources of data center cluster servers in conditions of a fractal structure of the network load by developing a dynamic algorithm for its distribution and balancing based on nonlinear dynamics methods.