The models of network control and resource management in stochastic environment are of great importance. The choice of models and methods depends on adequacy of identification and real measurements of traffic characteristics. Without robust theoretical preconditions network environment modeling will fail to simple simulation results. Therefore Internet as environment of control which is characterized by specific features including dynamism, hierarchy and adaptability is a topic of active fundamental researches. In the paper we consider Internet properties which can be presented by dynamic TCP model and stochastic character of observable fluctuations. Proposed approach is based on the analysis of experimental results, modeling, and analytical approximation of distribution function of stochastic intervals between fast recovery TCP phases.