Java Grid workflow management with fuzzy clustering approach

摘要

This paper presents a framework, based on Petri net and dynamic fuzzy clustering, to describe and to schedule task workflows in Grid environments. Computational Grids are intrinsically heterogeneous and dynamic systems, these features make difficult to build a resource scheduler that update itself automatically, following the resource evolution. In order to build a Grid scheduling system that automatically determines which grid resource should be used for each task, in this work we present the integration of a graphical tool to represent task workflows by Petri net, with a translation service to translate workflows in a meta language and a dynamic cluster service to schedule each task in the grid resource that best meets the specific task requirements.