Highly Dynamic Workflow Orchestration for Scientific Applications

Scientific applications executed with modern distributed technologies tend to be computed on various, dispersed resources, and workflows become a natural method of describing them. This method of the functional application decomposition allows for fully manual design only to a certain level of complexity. However, the scientific simulations usually expose non-trivial processing logic including multiple distinct processing elements connected with complicated control patterns. Thus an environment for semi-automated and assisted composition and orchestration of such application workflows is desired. This report describes the concept and design of such a platform that analyzes user requirements regarding application results and leads the user in the process of possible solution construction, dynamic refinement and execution. The tools forming the presented system use domain-specific knowledge and employ several levels of workflow abstractness in order to deliver the functionality in a more natural way for the human user. This work presents a real-life case study from the city traffic simulation domain, in order to introduce subsequent steps of workflow orchestration. It also supplies a discussion on existing similar systems for workflow orchestration and it finishes with a conclusions section. This research work is carried out under the FP6 Network of Excellence CoreGRID funded by the European Commission (Contract IST-2002004265).