DEVELOPMENT OF THE ANALYTICAL MODEL OF THE NGN INTELLIGENT SUPERSTRUCTURE WITH DECENTRALIZED CONTROL PRINCIPLE

Abstract

This article is dedicated to the development of the analytical model of the NGN intelligent superstructure with decentralized control principle of providing services based on the self-similarity of the stream of service requests. It is shown that the NGN concept appeared a long time ago, but to this day in Ukraine there are only several «islets» of NGN. The NGN supports Triple-Play Services functionality (voice, video and data transfer). It is noted that as of this day, the NGN uses the intelligent superstructure with Centralized Control Principle (ISCCP), which enables to provide the Triple-play services. Given the growing demand for triple-play services, the ISCCP is not always able to provide quality of service. To improve the quality of service, an NGN architecture with an intelligent superstructure that uses a decentralized control principle (ISDCP) is proposed, which eliminates the disadvantages of the ISCCP. It is suggested to evaluate the quality of functioning of the intelligent superstructure (in the scope of quality of service) on the basis of estimation of quality indicators, the probabilistic and temporal characteristics, using the analytical model of the NGN intelligent superstructure. This takes into account that the inbound stream of service requests in packet-switching networks has the features of self-similarity. This paper presents two approaches to the implementation of the ISDCP. In both the first and second cases, the services provided by the ISDCP are broken down into classes. Classification may be done, for example, by technology of providing services (or in accordance with QoS requirements). In addition, some services do not allow execution delays. These services should be grouped into a separate class. In the first case, each server contains a logic for servicing all classes of services (universal server). That is, each of the servers replicates all the features of the single server that functions in an intelligent superstructure with centralized control principle. The second approach involves the use of dedicated servers. In this case, it is possible to group the services according to the demand for certain types of services in the respective territory. For a certain set of classes of services, the logic of their rendering is placed in a separate service control node (server). Each set of classes has its own node. It is further suggested assuming the servers to be dedicated, and to consider the NGN architecture with this type of intelligent superstructure. It is stated that the development of the analytical model of the ISDCP needs to be based on the use of approaches of queuing theory and the model to be based on the application of the fractal Brownian motion. The ISDCP is presented in the form of a separate built-in network, and that one, in turn, in the form of a queue network (QN), indicating the main hypotheses and assumptions. It is shown that in this QN, each device will receive a grouped (aggregate) stream from several sources. It is thereto noted that when aggregating streams from several sources, if at least one of them features self-similarity, then the resulting aggregate stream will also feature selfsimilarity. Combining streams from traffic-generating sources described by a process with infinite dispersion results in self-similar network traffic that resembles the traffic described by the fractal Brownian motion model. In this case, the Hurst metrics of the aggregate stream will be the maximum of all aggregate stream metrics. An estimation of the intensity and the coefficient of stream dispersion based on the expressions of mathematical statistics is proposed. Using the fractal Brownian motion model and the results obtained by Norros, expressions are proposed to calculate the time of service of the request on the i-th server and the probability of loss of service request due to buffer overflow with incoming traffic featuring the effect of self-similarity. The proposed analytical model of the NGN intelligent superstructure with decentralized service control principle, taking into account the self-similarity of the stream of service requests, will allow calculating (predict) the values of probabilistic and temporal characteristics (quality of service indicators) more accurately, which will allow the developers to choose the necessary network equipment with more precision.