Ejecución eficiente de flujos de trabajos computacionales en entornos de Grid

Resumo

Among the different aspects that involve Grid Computing, efficient execution of workflows is focused by the present work.Wokflows do mirror human needs of complexity and Grid Computing technology starts to be mature for solving scientific problems, as it allows access to a great amount of resources. Nevertheless, due to its inherent nature, the Grid is not completely ready for executing many wokflow types in a reasonable time. In order to justify the adoption of this computing paradigm, the algorithms governing these workflows must be analyzed and optimizations must be implemented so their execution will gain efficiency. This work starts with an introduction to Grid Computing where its main components are explained and some significant infrastructures are described. As workflows are implemented by applications, this Chapter will serve also as an introduction to Grid porting, reviewing existing technologies at different levels. The simplest type of workflow, the high throughput applications, is studied at first. Here, a Fusion Physics application provided by CIEMAT7 needed to increase its amount of processed data. At the beginning, Grid Computing provided the needed efficiency. When this data need increased again, a chunk strategy was adopted obtaining good results. Complexity is increased at the following Chapter, where a bioinformatic application proposed by CNIO8 finds in the Grid the best solution to bypass the memory restrictions of a single machine. As the Grid execution of this application was not saving enough time, many optimization heuristics weres studied. From these, two have been chosen for implementation and further study. Additionally, this work provided a valid model to predict the execution time for a given set of starting conditions, also using the implemented heuristics.