Sandia Labs FY21 LDRD Annual Report


R&D 100 Award Winner Slycat: Providing scalable ensemble analysis and visualization.

Traditional scientific visualization systems provide tools to explore and compare no more than a handful of simulation results at once. However, evaluating computational models through sensitivity analysis, uncertainty quantification, and parameter studies all require creating collections of runs, known as ensembles. Ensemble sizes can exceed 10,000 runs, each with hundreds of changing variables, plus multimedia outputs. So as the scale of ensemble data grows, ensemble-specific tools are needed to understand the results. Slycat is a framework that addresses this capability gap by integrating data management, scalable analysis, abstract visual representations, and remote user interaction through a web-based interface. Slycat reduces data storage and movement costs by performing parallel analysis of the full data on the source cluster for the ensemble, then using the reduced-size analysis artifacts to model the ensemble on a Slycat server. Each Slycat model targets one or more specific result types, using multiple levels of abstraction to view the ensemble as a whole, results across variables, results between sets of runs, and components of individual runs. Slycat, inspired by an LDRD project led by PI Daniel Dunlavy in 2010, runs through a web-based interface designed for access-controlled collaboration between authenticated project members. Watch the YouTube video.

Members of the research team working on Slycat at Sandia. (Photo courtesy of Sandia)



Made with FlippingBook Ebook Creator