Network traffic characterization in a transaction based modeling environment

Workload characterization is difficult yet vital in the construction of realistic and useful performance models. Previously, a Poisson process has been used to represent the arrival of traffic to nodes in a network model. Traffic could be constructed as groups of related messages called transactions. Letting these transactions represent basic units of work, a hierarchy is described that allows for the construction of workloads representing traffic at the network user level. These user workloads are then decomposed into independent Markov modulated Poisson processes for use as input to analytic solution techniques based on the GI/G/m two-moment decomposition method