A Maya use case: adaptable scientific workflows with ADIOS for general relativistic astrophysics

Abstract

There are many challenges in analyzing and visualizing data from current cutting-edge general relativistic astrophysics simulations. Many of the associated tasks are time-consuming with large performance degradation due to the magnitude and complexity of the data. The Adaptable I/O System (ADIOS) is a componentization of the I/O layer that has demonstrated remarkable I/O performance improvements on applications running on leadership class machines while also offering new in-memory "staging" operations for transforming data in-situ. We have been incorporating ADIOS staging technologies into our Maya numerical relativity code based on Cactus infrastructure and Carpet mesh refinement. Incorporating ADIOS into the Maya code is the first step toward enabling a more adaptable Maya workflow. We provide descriptions how we intend to leverage FlexPath (an ADIOS transport method) to provide a richer user experience in real-time visualization and interactive steering.