Results of backbone internet channel informational flows mutual influence research, based on 15 minutes long free access MAWI archive [1] internet-traffic damps are discussed. Informational flows are divided into three classes: “Elephants” “Miles” and “Mice”, based on the traffic value transmitted by the flow. In accordance with the author's methodology, random sequences (RS) containing the time-ordered values of the number of packets and the amount of information transmitted during a given time interval, as well as the average amount of information transferred per packet were obtained for each of the selected classes of streams. It is demonstrated that, contrary to well-established notions, the fractal (self-similar) properties are possessed not by (RS) but their accumulated sums. Estimates of their Hurst indices are obtained. Regression models, that describes the dependencies of the Hurst indices of the accumulated amounts of the (RS) based on the amount of information and the number of packets transmitted by each of the selected user classes, which allow to evaluate the mutual influence of the information flows created by Mice, Mules and Elephants, on each other, were defined. It is demonstrated that the values of transmitted for 15 minutes intervals information are connected with deterministic linear dependencies. This demonstrates the possibility of developing mechanisms for channel load balancing, based on transfer information amount control of each of the selected classes of users depending on the current values of the Hurst indicators of accumulated sums of the (RS) that can be used to improve the quality of service for users of this channel. Algorithm example is provided.