Trustworthy and dynamic mobile task scheduling in data-intensive scientific workflow environments

There is an increasing demand for data-intensive applications in which scientists use scientific workflows to integrate together data management, analysis, simulation and visualization services over often voluminous complex and distributed scientific data and services. One major limitation of current scientific workflow models is that each workflow task is stationary, requiring a dataset to be transferred from its source host to a target host where a stationary task resides before a computation can be performed on the dataset. This limitation seriously impedes data-intensive applications since it can take an unbearable amount of time to transfer large amount of datasets from their sources to the host where a stationary task resides. In order to address this limitation, in this paper, we apply the idea of mobile agents to distributed scientific workflows. In contrast to stationary tasks, mobile tasks move from their home hosts towards datasets and perform computations on the dataset side. Since in dataintensive applications, it is often the case that the size of a mobile task is much smaller than the size of a dataset, our mobile-task approach can greatly reduce the network communication overhead. Since a mobile task might migrate across various administrative domains and get executed at multiple hosts, it is critically important to ensure the security of a mobile-task-based workflow system. The lack of effective access control model creates a security hole in distributed scientific workflows. In this paper, we address this problem using an itinerary-based access control model and a host visit scheduling algorithm that prevents arbitrary tasks from accessing and being executed on the current host.