Геометрический метод оперативного управления распределенным решением информационно-расчетных задач в вычислительных сетях

Abstract

Aim . Some of the main performance indicators of ACS application are the operational efficiency and stability of the control of the above mentioned systems. The wide application of computing techniques in ACS as well as the organization of computer networks on this basis stipulate the necessity of effective control of distributed computation processes to ensure the required level of operational efficiency and stability while solving the specified tasks. The existing methods used to organize the computation process (method of dynamic programming, branch and bound method, sequential synthesis, etc.) may turn out to be bulky or less accurate in certain situations. These methods help to find a solution in the mode of interactive choice of an optimal variant to organize a computation process, i.e. consecutive approach to the required result and do not allow getting an a priori estimation of the time of computation process in a network. Application of the specified methods when solving research tasks in the course of design of computer networks presents itself as quite difficult. This article offers the application of a geometrical method that allows estimating the minimum time necessary to solve the set of information-computing tasks as well as ensuring their optimal assignment in a computing system. Besides, the method allows finding a full set of possible variants for the organization of a computation process in a network with an a priori estimation of time of the decision for each variant. The principle of the method is to represent the sets of all possible distributions of tasks by workstations in form of a broken hypersurface. To solve the indicated task the criterion and conditions of the optimality of the time spent to solve informationcomputing tasks have been introduced. Results and conclusions. This article describes many variants of realization of a computation process for homogeneous and non-homogeneous computing environments. Solution algorithm for a homogeneous computing environment is quite simple and makes it possible to define a minimum time necessary for a computation operations. It is based on a geometrical representation of the distribution of tasks by workstations in form of the hyperplane constructed in orthonormal space whose basis vectors are computation capacities of workstations. Besides, the algorithm for homogeneous computing environment can be successfully used for an approximate estimation of the minimum time necessary to solve a set of tasks in a network, for non-homogeneous computing environment as well. Minimum time necessary to solve functionally different tasks in a non-homogeneous computing environment is defined using a piecewise linear hypersurface that slightly complicates the algorithm, though in general, with consideration of computation capabilities of moderns computers, it is still simply realized. The estimations carried out in the course of preliminary researches, allowed concluding about the application of a geometrical method in a computer network for a large amount of workstations and informationcomputing tasks. The possibility of an a-priori estimation of the minimum time necessary to solve a set of tasks in the computer network allows using the offered method to solve research tasks at the stage of design of a computer network to estimate such indicators as operational efficiency, reliability, stability and etc. The possibility of an aprioristic assessment of the minimum time of the solution of a complex of tasks in the computer network allows to use, offered in work, a method in the solution of research tasks at a design stage of the computer network for an assessment of her such indicators as efficiency, reliability, stability, etc.