A Generalized Bi-Objective Scheduling Algorithm for Batch-of-Tasks on Heterogeneous Computing Systems
The high energy consumption of data centers and its contribution towards greenhouse gases demand energy-efficient management of resources. Energy consumption of computing resources encourages the development of bi-objective scheduling algorithms optimizing the makespan of jobs and energy consumption...
Gardado en:
| Главные авторы: | , , , |
|---|---|
| Formato: | Статья |
| Idioma: | English |
| Publicado: |
International Institute for General Systems Studies
2024
|
| Темы: | |
| Acceso en liña: | https://dspace.ncfu.ru/handle/123456789/28756 |
| Метки: |
Engadir etiqueta
Sen Etiquetas, Sexa o primeiro en etiquetar este rexistro!
|
| Краткое описание: | The high energy consumption of data centers and its contribution towards greenhouse gases demand energy-efficient management of resources. Energy consumption of computing resources encourages the development of bi-objective scheduling algorithms optimizing the makespan of jobs and energy consumption of computing resources. In general, the problem of job scheduling and bi-objective optimization falls in the NP-complete combinatorial optimization problem category. To address the bi-objective scheduling problem, a generalized bi-objective scheduling algorithm (Z*) for Batch-of-Tasks (BoT) applications on the Heterogeneous Computing System (HCS) has been proposed. The BoT represents the set of independent tasks from multiple applications, and the HCS represents the computational environment consisting of processors with different frequencies. To schedule tasks, the Z* algorithm takes decisions using the optimization function of energy consumption and completion time of tasks based on the given weights. The weight could be fractional or integer, so the Z* algorithm represents a set of different algorithms. The proposed algorithm is beneficial for cloud data centers/service-oriented computing to execute customer jobs based on the demand, whether the customer needs high throughput or low cost of execution. |
|---|