A Matching Model for MAP-2 using Moments of the Counting Process

In the context of network capacity planning and management, traffic matrix (TM) estimation plays a crucial role, as it provides the means for computing the volume of traffic that flows between all pairs of nodes in a network. Since the TM estimation problem was first addressed in [13], various techniques for its solution have been introduced based on observed traffic counts on every link in the network. In operational IP networks, these link counts can be obtained from readily available link load measurement data provided by SNMP. If a TM is obtained systematically over a number of fixed time periods, the components of these traffic matrices form a counting process. In existing statistical approaches for TM estimation, this counting process has been modelled by conventional traffic models such as Poisson [13], and Gaussian [5], since these traffic models have attractive theoretical properties. These approaches, however, have been shown to have limited accuracy and cannot be generally applied to practical IP networks [11]. Thus, increasingly, there is a need for new methods that make more accurate assumptions about the traffic characteristics of the flows between OD pairs.