A Dynamic Load Balancing Method for Data Centers with Fractal Properties of Network Traffic

This paper proposes an approach to developing and studying a load balancing system for data centers with the fractal properties of network traffic. Due to such properties, it is possible to forecast reliably the occurrence of bursts and declines of network traffic on separate time intervals and periods with possible overloads on servers and network equipment. Hence, it is possible to develop methods for effective planning and distribution of tasks within a data center to ensure statistically uniform loading of its functional elements and avoid overloads. The dynamic load balancing method is based on the statistical analysis of input network traffic (distribution density, autocorrelation function, spectral density, and fractality level).