ANALYTICAL CONSTRUCTION OF OPTIMUM CONTROL COMPUTER NETWORKS

Abstract

This article is about developing a computer network control system. In order to synthesize the control system of a complex technical object - a computer network - it is advisable to use optimization methods based on the minimization of control quality functionals. The quality of the computer network control system depends on the type of optimization target functionality: speed, accuracy, amount of additional control actions (information signals), etc. The purpose of the work is to compare the optimal control systems for analytical design by different criteria. Analytical design of a computer network control system, optimized by the Letov-Kalman criterion and generalized work by A. Krasovsky showed the advantages of the latter method. The difficulties of the analytical method of the Letov-Kalman criterion are related to the large volume of necessary calculations. The problem of analytic construction by the criterion of generalized work is reduced to solving a linear matrix equation, while analytical construction by the Letov-Kalman criterion requires solving a nonlinear matrix equation. The complexity of the solution on the basis of the criterion of generalized work is significantly (approximately n / α times) less than in the synthesis by the Letov-Kalman method. This difference is particularly noticeable for high-dimensional objects such as computer networks. The paper presents the comparative characteristics of the required computing power for a control system designed according to the Letov-Kalman criterion and generalized work. Generalized work functionality is shown to have an advantage in computing power of approximately 10⁴ op / s with the same number of network monitoring parameters. The dependences of the required computing power on the degree of sparse matrix of the network state matrix (for different modes of operation) are also analyzed.