It is pertinent to develop a set of simple parameters that can best characterize the variability and the statistical correlations of the packet arrival process. These parameters are then used by the network to allocate its resources among the different users in order to avoid congestion and maintain a predefined quality of service (QOS) for each user. The traffic characterization and modeling of the real-time bursty traffic, mainly voice and video traffic sources, using statistical methods are described. It is shown that the variability of the variance of the sum of consecutive packet interarrival times leads to significant queuing delays and therefore is a major cause of congestion in broadband networks.< >