Autonomic execution of computational workflows

This paper describes the application of an autonomic paradigm to manage the complexity of software systems such as computational workflows. To demonstrate our approach, the workflow and the services comprising it are treated as managed resources controlled by hierarchically organized autonomic managers. By applying service-oriented software engineering principles, in particular enterprise integration patterns, we have developed a scalable, agile, self-healing environment for execution of dynamic, data-driven workflows which are capable of assuring scientific fidelity despite unavoidable faults and without human intervention.