Loss ratio simulation method of MMPP/D/1/K using statistical techniques

For network planning and the operation of video streaming multicast services, it is very important to grasp the characteristics of packet loss against traffic type and buffer size because multicast uses UDP, not TCP, so packet recovery is not assured in this layer. This is possible by using Monte Carlo simulations (MCS), one of the most effective techniques. Although approximate methods have also been proposed as alternatives, in most cases, MCS results are provided to validate the proposal. However, there is no obvious way of determining when MCS should be terminated. This paper proposes a new obvious termination method for embedding in the MMPP/D/1/K MCS program that focuses on loss interval; it employs statistical techniques that allow loss ratios to be accurately estimated in the shortest possible time. The simulation is terminated when the χ²-statistic becomes less than or equal to a critical value, and the loss ratio is calculated from the number of all losses up to that time. A method of determining the parameters needed to calculate the χ²-statistic is also proposed to reduce the calculation time in MCS. To evaluate the proposed method, confidence intervals are obtained by applying the proposed method to six cases. Long-term loss ratios for the six cases are within 95 percent confidence intervals. 95 percent confidence intervals for the six cases are near approximate loss ratios. Embedding the proposed method in the MCS program imposes a time penalty of only 3.6 percent at most. The proposed method is thus very useful and usable.