Autonomic Execution of Computational Workflows

This paper describes the application of anautonomic paradigm to manage the complexity of softwaresystems such as computational workflows. To demonstrate ourapproach, the workflow and the services comprising it aretreated as managed resources controlled by hierarchicallyorganized autonomic managers. By applying service-orientedsoftware engineering principles, in particular enterpriseintegration patterns, we have developed a scalable, agile, selfhealingenvironment for execution of dynamic, data-drivenworkflows which are capable of assuring scientific fidelitydespite unavoidable faults and without human intervention