Анализ очереди типа G/G/1 для программно-конфигурируемых сетей на основе openflow

Abstract

The SDN imposes requirements on manufacturers of infocommunication equipment related to the support of new scenarios for working with network applications, for example, with cloud technologies, with the ability to transmit large amounts of traffic over long distances. All this leads to the need to analyze the features of the SDN networks, assess their performance both at the design stage and in the process of operation. Quality of service (QoS) parameters in networks, such as delay and delay variation, can be used as characteristics to evaluate the quality of operation. An analysis of the functioning of a software-defined network using the M/G/1 and G/G/1 systems was made. Solutions based on two approaches are given, when the truncated normal distribution and the hyperexponential distribution are used as an example of an arbitrary distribution. When analyzing the G/G/1 system, an assumption was made about the mutual independence of flows entering the system and the absence of correlations within the sequences of time intervals between requests and processing times of requests. In the work, an expression for the probability density of processing times of applications in the systems M/N/1, H2/N/1 and H2/H2/1, analytical expressions for estimating the average values of the delay time of applications in the systems M/N/1, H2/N /1 and H2/H2/1 for software defined networks. The paper presents the result of comparing the average values of estimates of the delay time of applications in the systems M/N/1, H2/N/1 and H2/H2/1 for the SDN. An analysis of the dependence of the average delay time of a request in the system on the load factor in the queuing systems under consideration is given.