Analysis and traffic characterization of a wide area network

The characteristics and dynamics of wide area network (WAN) traffic are discussed. The real-time experimental prototype system for monitoring and analyzing traffic allows the monitoring of different parts of the network for extended periods of time without a need for storing and processing the voluminous amounts of data that are captured during the monitoring periods. The results of the analysis show that the inter-site traffic follows a certain pattern that depends on the time of the day and the day of the week. The traffic between two sites is found to be reasonably symmetric, except for those sites designated as backup sites. WAN traffic is highly dynamic and up to 25:1 peak-to-low ratios are observed over a one-week monitoring period. Using the coefficient of variation as a measure for the traffic burstiness, a 1.5 degree of burstiness is maintained during the working hours. This moderate degree of burstiness can be attributed to two factors: the smoothing effect that results from aggregating a number of bursty sources, and the use of 1.5 Mb/s T1 links between sites that usually reshapes the traffic and changes its characteristics (i.e., reduces its burstiness). A two-state Markov modulated Poisson process (MMPP) model is developed to represent the correlation between packets, as well as a benchmark for the design and performance tuning of future wide area networks