Strykhaliuk B. Structural and functional synthesis of heterogeneous service-oriented telecommunication networks.

The thesis is devoted to solving of actual scientific problem of optimal structural and functional synthesis of heterogeneous networks based on Cloud-technologies in the process of providing a large range of services under the influence of external and internal dynamic structural and parametric effects. In this paper we propose a model of hyperbolic Ricci flow with conformal mapping of structural set of vertices of the cloud-based network within Riemannian metrics virtual coordinate system of network nodes, which resulted in availability improving of the cloud network nodes up to 17%. We propose a model based on Lagrangian optimization, that makes possible to carry out a negotiated solution of the nonlinear problem of routing in the СSDP system and virtual servers migration tasks by adapting the probability distribution of the modeling parameters of the previous virtualized resources of cloud systems to ensure the successful request execution at not less than 99,9% probability. Using the proposed solution of the optimization problem in adaptation to the average allowable delay based on service priorities allows, firstly, considering the functional features of service compositions and their mutual influence on the process of virtual service structures synthesis more solidly. Secondly, to ensure the required quality of service for each composite application that is implemented being based on the proposed architecture. Third, to propose a multicast transmission method for data flows in distributed data-center networks to increase the usable capacity of such systems with a 10 Gbit/s basic throughput up to 35-42% and reduce the probability of re-transmission. Further development of the multicast transmission model for Fat-tree, VL2, and B-Cube topologies was carried out based on the introduction of multicast backup plane, thus reducing by half the traffic overhead caused by traffic footprint that occurs due to loss of queries in a distributed data-center network. Based on these studies, a service optimization model for СSDP architecture is proposed, which helped to improve the overall performance of the heterogeneous service platform in terms of migration of virtual machines up to 15% in the case of virtual machines migration containment for the resource utilization values around 0.3, and up to 10% in the case of resources utilization coefficintes, greater than 0.8.