Minukhin S. Models, methods, information technologies of job batch scheduling in distributed computing systems

The thesis is devoted to solving the scientific and practical problem of increasing the efficiency of distributed computing systems (DCS) maintenance on the basis of models, methods and information techniques of job batch scheduling using a single theoretical approach to the solution of NP-complete problems of discrete optimization and graph theory. The concept, principles and propositions of the job scheduling and its implementation model, as well as the mathematical scheduling model in the form of the minimal cover problem solution are suggested. The model of economic resource is developed based on the determination of the equilibrium market price of computing resources of distributed computing systems in the conditions of a double auction. The methods of solving the problem of minimal cover with a low time complexity are devised. The methods of ensuring the quality of service at local DCS resources are developed within the scope of the total tardiness problem, reduced to constructing the shortest Hamiltonian path in a graph using ranking approach and the rules of dominance and to the problem of feasible tasks schedule based on the methods of scaling the frequency and the voltage of processors within a computing cluster. The methods of constructing feasible schedule of tasks for processors of computing clusters of various architecture providing a balance of their load are developed. Information technologies to implement a twolevel scheduling model of job batch at the level of grid segment and computing clusters of DCS are improved.