| Abstrakt: |
This paper describes the mathematical formulation of the problem of task flow scheduling in GRID – systems with inalienable resources, taking into account the power of the node and the power of the problem, as key factors affecting system performance. There are four methods for solving it. One for easily parallel tasks, the other three for sequential tasks. Described client - server architectural model for building software for distributed computing and tasks that require high computing power of the system. The expediency of comparing the efficiency of the proposed methods with the existing FCFS method, as well – known and widely used in various GRID – systems, is substantiated. Describes the test results, which showed that all new methods give a better result than FCFS. [ABSTRACT FROM AUTHOR] |
