Tkach M. Development and research of efficiency of multiprocessor systems with an expanding area of computing

The dissertation is devoted to solving actual scientific and technical problem of development of theoretical and practical bases of creation of high-performance multiprocessor computing systems with distributed computing field based on use of modern functional modules of mass industrial production, which ensure the solution of problems requiring large amount of CPU time. The concept of research productivity indicators for a high-performance multi-processor system with distributed calculation field has been proposed. The ability to evaluate the effectiveness of a multiprocessor cluster system through its parameters, which allows to assess the appropriateness of the selection of hardware of the developed system, has been implemented. The concept of influence of the network interfaces on the performance characteristics of multiprocessor systems with a distributed field of computing has been proposed and investigated. The basic operation modes of the network interface in such systems were analyzed and their impact on the slowdown of the computation was determined. This approach allowed to identify ways to reduce the deceleration rate calculations in multiprocessor systems. Theoretical and practical aspects of the prospects for using modern communication technologies in multi-processor cluster systems with distributed computing area have been investigated. A comparative analysis of the performance of a multiprocessor cluster system evaluations using various types of network technologies was fulfilled. The main components of the parallel design technology algorithms based on numerical and analytical approach have been proposed and implemented. It is shown that the proposed approach assumes the highest degree of calculations' vectorization, determines the maximum parallel algorithmic form and allows to increase the efficiency, productivity and computing performance. A complex of applied programs for solving problems of metallurgical production has been developed. The main principles of data visualization and management were proposed, and basing on them, one can obtain an adequate picture of the phenomenon under investigation.