Grid computing has become in recent years an important alternative to traditional parallel computing by providing computational capabilities at large scales.
On the other hand, complex networks have shown to develop non-trivial topological features which can be adapted to the optimization of a wide range of problems.
The present study uses this approach for the reduction of applications execution time in a Grid infrastructure. In this work, we focused on those characterized by parametric sweeps. The way to achieve this task is by applying the preferential attachment technique with a small modification: new nodes are added to the evolving graph with a probability proportional not only to the target node degree but also to the efficiency of resources involved.
As a result, the application becomes self-adaptive to the infrastructure by dynamically obtaining resource rankings and classifying them according to their efficiency.