High performance Computers allow researchers to greatly accelerate their work. However, traditional scientific workflows force researchers to follow a sequential set of steps filled with dead times in between (model preparation/job submission/wait/data transfers/wait/analysis/wait/visualization) to gain scientific insights. This gets worse due to integration of many different local and remote tools for job preparation and analysis. In this paper, we present the design, implementation, testing, and validation of an integrated scientific workflow for PELE–a Monte Carlo simulation software used in molecular modeling.
We used formal user-centered design methods from the conceptualization to the validation stages. Our workflow integrates job configuration, simulation steering, real time analysis, and visualization through an intuitive multiplatform user interface. A major contribution of our design is offering intuitive tools to inspect the evolution of the simulation by providing in-situ analysis and visualization of key metrics and its molecular systems. After user evaluations, we found these tools were the most useful so users could rapidly decide whether to pause, stop, or change job configurations and steer the simulation on the fly, reportedly reducing the time-to-solution (both human plus computing time and effort) up to an order of magnitude.