Isolation Levels for Data Sharing in Large-Scale Scientific Workflows

Scientists can benefit from Grid and Cloud infrastructures to face the increasing need to share scientific data and execute data-intensive workflows at a large scale. However, these workflows are creating more and more challenging problems in the automation of data management during execution. Existing workflow management systems focus on how data is stored, transfered and on data provenance. However they lack in managing isolation during the execution of tasks of the same or different workflows that read/update shared data. In this scope, we propose three isolation levels taking into account data provenance and multiversioning. In the best of our knowledge this is the first proposal in such context.