Вісники та науково-технічні збірники, журнали
Permanent URI for this communityhttps://ena.lpnu.ua/handle/ntb/12
Browse
Search Results
Item Расчет характеристик самоподобного трафика, аппроксимируемого распределением Парето(Видавництво Львівської політехніки, 2017-03-28) Lozhkovskyi, A. G.; Levenberg, Ye. V.; Ложковский, А. Г.; Левенберг, Е. В.; O. S. Popov Odessa national academy of telecommunications; Одесская национальна академия связи им. А. С. ПоповаThe model of self-similar traffic is widely used, but some tasks of assessing the quality of service in the packet network are unresolved. It is known that in the presence of self-similarity properties in the incoming stream of requirements with increasing load intensity ρ, the quality of service characteristics deteriorate, but not as much as is assumed by the Norros method. The discrepancy between the results of modeling and estimates obtained by the Norros method is hundreds of percent. It is obvious that the estimate of Norros formula is significantly overestimated, which requires finding a more accurate solution. A method is proposed for increasing the accuracy of calculating the quality characteristics of servicing self-similar traffic due to a more accurate determination of the Hurst coefficient as a function of the parameter of the Pareto distribution form. For the self-similar traffic described by the Pareto distribution, a new formula for calculating the Hurst coefficient is obtained. In this case, the calculation of the quality of service characteristics can be performed on the basis of the Norros formula. The proposed method allows us to calculate the quality of service characteristics of self-similar traffic described by the Pareto distribution in a single-channel system with discrete time of packets service is much simpler. This simplicity is explained by the fact that in order to calculate the self-similarity coefficient of Hurst, you only need to know the parameter a of the Pareto distribution form and you do not have to calculate for this traffic in a very complicated way, for example, using absolute least-squares methods or R/S-statistics.