Кафедри
Постійне посилання на розділhttps://repository.kpi.kharkov.ua/handle/KhPI-Press/35393
Переглянути
1 результатів
Результати пошуку
Документ Composite application distribution methods modeling(Національний технічний університет "Харківський політехнічний інститут", 2018) Bulba, S. S.The subject of consideration are algorithms for optimal distribution of existing pool of computing resources between composite applications and algorithm of utilization of resources on computing blocks. The purposeof the article is to analyze the results of simulation and mathematical modeling of the resource allocation process between composite applications, depending on the distribution option. The efficiency of existing dynamic planning algorithms that are related to the greedy algorithm class is considered. They find a locally optimal solution at each step. The boundary of effective planning of algorithms based on clustering approach is revealed. The efficiency of using ant colony optimization algorithm and algorithms of cluster approach using ant colony optimization algorithm is shown. The simulation of the distribution of the composite application is carried out, depending on the complexity of the graph construction. The dependence of the execution time of the composite application on utilization of resources on the calculated blocks is obtained. Using the resource utilization function, the quality of the distribution of composite application resources is analyzed, depending on the amount of data transferred to the calculations. Data on the quality of resource allocation is obtained, depending on such parameters as the time of implementation of the composite application, the volume of transmitted data, the complexity of the graph construction. A method for choosing the optimal resource allocation algorithm between composite applications depending on the listed parameters is proposed. This will allow you to quickly dispose of distributed computing blocks that are occupied by calculating a distributed task, which will speed up the computation of distributed tasks on an existing pool of computing blocks.