The present article describes simulation results for single queue P/M/1/K model. Simulink has been used to create this model and evaluate its performance. The main purpose of the research made is to optimize costs of resources allocation for TCP traffic type at the same time striving to achieve the best packet loss probability. The optimization was made with dynamic programming method by using Bellman algorithm for simulated traffic with different self-similarity (Hurst) parameter and utilization coefficient values. The results of this work indicate that Hurst parameter affects greatly the costs of resources to be allocated. With estimation of Hurst parameter for network traffic these results can be used to estimate buffer memory capacity and ensure the specified value of packet loss probability. The research made also suggests to increase buffer memory capacity rather than channel bandwidth, whenever it is possible.