This series of papers deals with developing of methods, models and tools for control of complex distributed computing systems providing services to users under conflict and uncertainty. Distributed users-oriented systems are complex and highly important subject of research. The most important application areas of application of the theoretical results are the Internet and modern multiprocessor computing environments , such as Grid and Cloud. In the first case the service is network bandwidth, in the second - CPU resources.
In this work we apply a unified game theoretic approach to the analysis of such systems in which users experienced with situations of competition for system resources, conflict interaction with malicious users and targeted actions aimed at disruption of the functioning of the system – Denial of Service Attacks.
High complexity of the processes occurring in these systems stimulates interest to analytical models and methods that would investigate the behavior of the whole system. These research methods are methods of the theory of differential games, optimization methods of resource allocation in heterogeneous multiprocessor systems, methods of automatic generation of programs.
The paper presents the main results of the following:
• A set of analytical fluid models of networks with protocols Transmission control protocol ( Transmission Control Protocol - TCP ) and User Datagram Protocol ( User Datagram Protocol - UDP). Investigated and proved the correctness of the application of the theory of dynamic games to development of network control models;
• Constructed game of N users competing for network resources. There are found the conditions of existence of Nash, Pareto and evolutionarily stable equilibriums;
• Constructed fluid deterministic model of TCP network under denial of service attack. There are found conditions of existence of solutions, and analytical expression for vulnerability of system to different types of attacks;
• Built algebrodynamical execution model for multithreaded programs and videographical platforms. On the basis of the constructed models created tools automate the conversion of programs to improve their effectiveness runtime;
• Developed tools rewriting rules used for the automation of parallelization and optimization programs, in particular in Fortran. Implemented conversion to map from sequential to parallel multithreaded program, as well as improve the efficiency of parallel programs through better use of synchronization constructs and location of objects in memory;
• Developed automated conversion programs for the transition from sequential programs to parallel programs for video graphics accelerators using platform NVidia CUDA, as well as to improve the effectiveness of programs for video graphics accelerators;
• The analytical model of the execution time for the video graphics accelerators, using information about the structure of the program , as well as time to perform basic operations videographic devices. On the basis of the constructed models implemented instrumental system simulation gpusim using open system GridSim.
Results of studies that are nominated for the award are published in 19 publications, including 2 chapters in the book publishing house Springer, 4 publications in journals that are maintained in the database Scopus, a reports were presented on 3 foreign conferences. General topics of research by the authors were published in 44 works .