Abstract (EN):
Many scheduling algorithms for workflow applications target a single application at each time. The degree of parallelism of the application depends on the workflow structure, and for a task parallelism approach, the degree of parallelism will determine the degree of utilization of the resources reserved for the application. As today¿s computing nodes have several cores and considerable computing power, the single application approach may lead to higher computational costs when considering the energy spent. A plausible alternative is to share resources among applications, avoiding static reservation of resources, so that the energy spent by a job is minimized. In this chapter, we demonstrate the advantage of resource sharing and present a review of algorithms that have been successfully proposed for concurrent workflow scheduling. Challenges for grid resource managers, to accomplish an energy efficient resource sharing, are devised. Moreover, a review of the economic models for resource management and pricing, together with a case study, are also presented.
Idioma:
Inglês
Tipo (Avaliação Docente):
Científica
Contacto:
jbarbosa@fe.up.pt
Nº de páginas:
25