Large-scale scienti c computing often relies on intensive CPU tasks chained through a workflow running on a high-performance environment (HPC). In this context, scientists model their workflows for later submission on dedicated HPC making use of Scienti c Workflow Management Systems (SWfMS), which may improve data management on scienti c workflows. Unfortunately, when SWfMSs execute experiments as black-boxes, scientists typically nd ways of tracking the execution, i.e following the evolution of computations by opening and browsing les commonly spread over a distributed architecture. To track the convergence of a given experiment, scientists even try to stage out some data to visualize the workflow execution. However, such process is complex and error prone, mainly because the user has to \guess" the context of this partial result generation once les and their metadata are not connected. So that, this scenario hides potential issues given the interface between scientists; workflows models; experiment data; and the execution environment. In this work, we propose a scienti c portal, which allows users to manage the execution of large-scale scienti c workflows based on runtime provenance queries. Proteus provides a suitable architecture that supports and integrates with new applications aimed to take advantage of unbound access to experiment's provenance and execution environment. In order to evaluate the proposed architecture, we implemented and integrated a rst Proteus' aplication, dedicated to enhance experiment's visualization. This application integrates with Proteus promoting provenance visualization based on provenance queries. Therefore, Proteus is evaluated while providing support to port new applications, as well as providing partial visual analysis of a Uncertainty Quanti cation experiment enriched by provenance.
.png)
.png)
