IP Network Performance Measuring Based on I/O Correlation Model

Abstract

Network performance measuring provides an effective way of collecting first-hand’s data for network behavior analysis. It is also the foundation for high performance protocol analysis, network equipment development, networks plan and management, and applications. Japanese scientists have first found that network traffic manifests high auto-correlation from long-range dependence. Based on analysis of network behavior and traffic between its input and output (I/O), the study presents a measuring model based on the I/O correlation for network performance testing. The proposed method is applied to practice. Experimental results demonstrate that network traffic exhibits asymmetric volatility in both short-range and long-range observations. I/O correlations vary in the short-range observation, while similarity and autocorrelation from long-range dependence is implied. The measuring results from long-range observation reflect in accord with actual states of the network at different periods. The results also conclude that network traffic characterizes itself as similarity, autocorrelation and asymmetric volatility from long-range dependence, and with varying degrees in short-range correlation. The proposed method is shown its potential practicality.