MACRO SYNTHESIS OF WIDE AREA COMPUTER NETWORKS AT NONLINEAR COSTS

Abstract

In practice, simple solutions of quick preliminary estimation of basic characteristics of a wide area computer network sometimes are needed. To achieve this goal, the backbone data transfer subnet and server set of computer networks are opportune to be examined. Considering the nonlinear continuous dependence of the costs of channels, routers, and servers on their performance and using the J.R. Jackson’s partitioning theorem, an analytical model for these components of the network is defined. Based on this model, two optimization problems are formulated and examined: minimizing the average retention time of data processing requests and minimizing the summary cost of servers, channels and routers of the network. For both problems, numerical solution algorithms are proposed. These algorithms allow obtaining the expected solutions in continuous form, the respective performances being expressed by positive real numbers. Since the allowed performances of the mentioned entities are discrete, additional adjustments to the solutions in question, depending of the case, may be necessary. For such adjustments, two algorithms are proposed. One of them solves the problem by reducing it to that of the backpack. Another solves the problem in an heuristic manner based on the L. Kleinrock’s resource concentration rule.