Wyniki wyszukiwania - BazTech

Ograniczanie wyników

2 Image Processing & Communications

Znaleziono wyników: 2

Liczba wyników na stronie

Wyniki wyszukiwania

Sortuj według:

Ogranicz wyniki do:

Processing Astronomical Image Mosaic Workflows With An Expert Broker In Cloud Computing

Prado R. P., García-Galán S., López L. R., Expósito J. E., Reche R. R.

Image Processing & Communications

2014

Vol. 19, no. 4

5--20

Montage image engine is an astronomical tool created by NASA’s Earth Sciences Technology Office to obtain mosaics of the sky by the processing of multiple images from diverse regions. The associated computational processes involve the recalculation of the images geometry, the re-projection of the rotation and scale, the homogenization of the background emission and the combination of all images in a standardized format to show a final mosaic. These processes are highly computing demanding and structured in the form of workflows. A workflow is a set of individual jobs that allow the parallelization of the workload to be executed in distributed systems and thus, to reduce its finish time. Cloud computing is a distributed computing platform based on the provision of computing resources in the form of services becoming more and more required to perform large scale simulations in many science applications. Nevertheless, a computational cloud is a dynamic environment where resources capabilities can change on the fly depending on the networks demands. Therefore, flexible strategies to distribute workload among the different resources are necessary. In this work, the consideration of fuzzy rule-based systems as local brokers in cloud computing is proposed to speed up the execution of the Montage workflows. Simulations of the expert broker using synthetic workflows obtained from real systems considering diverse sets of jobs are conducted. Results show that the proposal is able to significantly reduce makespan in comparison to well-known scheduling strategies in distributed systems and in this way, to offer an efficient solution to accelerate the processing of astronomical image mosaic workflows.

Uncertainty-tolerant scheduling strategies for grid computing: knowledge-based techniques with bio-inspired learning

García-Galán S, Prado R, Expósito J E

Image Processing & Communications

2013

Vol. 18, no. 1

37--44

Nowadays, diverse areas in science as high energy physics, astronomy or climate research are increasingly relying on experimental studies addressed with hard computing simulations that cannot be faced with traditional distributed systems. In this context, grid computing has emerged as the new generation computing platform based on the large-scale cooperation of resources. Furthermore, the use of grid computing has also been extended to several technology, engineering or economy areas such as financial services and construction engineering that demand high computer capabilities. Nevertheless, a major issue in the sharing of resources is the scheduling problem in a high-dynamic and uncertain environment where resources may become available, inactive or reserved over time according to local policies or systems failures. In this paper, a review of scheduling strategies dealing with uncertainty in systems information by the application of techniques such as fuzzy logic, neural networks or evolutionary algorithms is presented. Furthermore, this work is centered on the study of scheduling strategies based on fuzzy rulebased systems given their flexibility and ability to adapt to changes in grid systems. These knowledge-based strategies are founded on a fuzzy characterization of the system state and the application of the scheduler knowledge in the form of fuzzy rules to cope with the imprecise environment. Obtaining good rules also arises as a challenging problem. Hence, the main learning methods that allow the improvement and adaptation of the expert schedulers are introduced.