Tkachenko O. Methodology for constructing control systems for future networks based on queuing theory

The thesis is devoted to the solution of scientific and technical problems of building future network control systems. Analyzed are the methods for constructing systems for managing telecommunications networks, methods for constructing systems that are capable of self-adapting, self-learning and self-organization. The requirements for the NGN management systems are formulated, and the features of the multiservice network management are considered. It is proposed to use the principle for designing future control systems by networks of the future, the main idea of which is to define a law or optimal control strategy that minimizes or maximizes some set of quality criteria. The control law expresses the control action as a function of the coordinates of the state of an object, which results in a system with feedback. This technique simplifies the design of non-stationary and multidimensional systems and allows you to set the structure of the optimal system. Comparison of control systems algorithms is performed and it is proposed to use systems with automatic search for solutions as the optimal algorithm by providing them with the properties of systems with accumulated experience and using the method of searching for optimal choices. The method of constructing a control system based on a client-server architecture, according to which tasks in control systems are divided between clients and servers, has been further developed. In the platforms of future network management systems, it is proposed to implement a three-client-server architecture. Its main advantage is scalability, but, in addition, it can significantly reduce network traffic. A method of multicriteria optimization of control systems for networks of the future on the basis of queuing theory has been developed. The dependences of the delay in obtaining information on the intensity of demand, on speed, as well as on the method of service are obtained. These data minimize latency under various conditions. A method has been developed for constructing a simulation model of a control object using the principle of a discrete situational network (DSN), which, when modeling a control system, changes the structure of this system, the temporal characteristics of certain subsystems, and the order of operation.