As simulations grow in scale, the amount of data produced grows correspondingly, challenging existing strategies for I/O, analysis, and visualization. A large body of work has been devoted to addressing these challenges for structured data; however, the bulk of existing work targeting particle data has focused on the last two challenges, largely ignoring I/O. As such, particle simulations typically output data in a format which is convenient for the simulation to write, which must then be converted to a visualization or analysis optimized format in a time consuming post-process. In this talk I will discuss ongoing work to address these challenges holistically, with the aim of developing an I/O strategy and data layout for particle simulations which can be written efficiently at scale and used directly for visualization and analysis.
Posted by: Steve Petruzza