Ivan D. The data flow control efficiency increasing in backbone networks with MultiProtocol Label Switching.

The thesis is devoted to research the effectiveness of flow management in networks with Multiprotocol Label Switching, (MPLS), elaboration of methods, hardware, software and algorithmic tools to ensure its improvement. The effectiveness of flows' managing in such networks is determined simultaneously by several interrelated variables: the performance of network system, the utilization index of existing network resources, set in line to the used configuration methods of information flows and network systems' equipment that enable its sustainable operation in conditions close to full load. The paper presents methods for flow control and configuration of MPLS networking systems on the basis of the information flow distribution forecasting results by network topology analyzing with methods of graph theory and matrix calculus. Existing methods for the flow managing not offers a solution of hyper-aggregation problems arising from overload of the tracks formed by the vector-distance criterion, excluding the loading status of all potentially available channels to establish a route that leads to the necessity of solving the problem of flow management to improve the efficiency of all potential resources network bandwidth utilization, reducing the expense of the probability of information flow blocking and allow their dynamic re-configuration. The work results with further development of modeling techniques and configuration management systems' flows in MPLS-based streaming load forecasting by analyzing the topology of the network system. This is achieved by the formation of clusters of network structural system that configures and balances information flows in accordance with the projected needs of the network, which ensures efficiency streaming network resources utilization. A routing algorithms that take into account projected workload of the network system channels are obtained by analyzing its topology and graph theory methods using matrix calculus, conducted their modeling and experimental verification that shows the flow management system performance stability increasing and network performance indexes on amount of 20-30% in a general overload of the network structure, due to more efficient resource allocation of network bandwidth. Algorithmic complexity of the developed algorithms are polynomial and corresponding to the complexity of the routing algorithms used in modern transportation networks, which enables flexible dynamic reconfiguration of information flow. Described in this thesis algorithms and mechanisms constitute a synthesized adaptive foundation for the routing processes optimization. They are built and implemented using known classical algorithmic solutions. These mechanisms allow reducing the set of solutions to the desired. With them the best workflow management option can be selected, for instance: with minimum possible delay; or maximal possible probability/reliability of the workflow establishment; with the network resistance to the overloads and, consequently, network traffic alignment. The thesis proposed a hybrid adaptive low-level architecture at ISO/OSI physical and channel levels for cluster of integrated optical MPLS access network based on optical transmission of information using multi-phase channel code division multiplexing method, which has greatly enhanced the efficiency of the capacity distribution of optical channels at high-speed access infrastructure by using a large number of independently multiplexing code channels formed at the same optical carrier wavelength.