Predicting of Data Center Cluster Traffic

Abstract

The paper proposes a method for predicting of Data Center cluster traffic. Such prediction is based on the construction of the traffic probability density function. The method uses the expansion of the sample size due to the continuous majorant of the distribution function and, for small samples, gives a more accurate and stable evaluation than the existing methods. The method is more effective in analyzing a traffic with long-term pulsations, as well as long-term dependent traffic.