James Ahrens

Extreme scale scientific simulations are leading a charge to exascale computation, and data analytics runs the risk of being a bottleneck to scientific discovery. Due to power and I/O constraints, we expect in situ visualization and analysis will be a critical component of these workflows. Options for extreme scale data analysis are often presented as a stark contrast: write large files to disk for interactive, exploratory analysis, or perform in situ analysis to save detailed data about phenomena that a scientists knows about in advance. We present a novel framework for a third option - a highly interactive, image-based approach that promotes exploration of simulation results, and is easily accessed through extensions to widely used open source tools. This in situ approach supports interactive exploration of a wide range of results, while still significantly reducing data movement and storage.

Abstract The considerable interest in the high performance computing (HPC) community regarding analyzing and visualization data without first writing to disk, i. e., in situ processing, is due to several factors. First is an I/O cost savings, where data is analyzed/visualized while being generated, without first storing to a filesystem. Second is the potential for increased accuracy, where fine temporal sampling of transient analysis might expose some complex behavior missed in coarse temporal sampling. Third is the ability to use all available resources, CPU's and accelerators, in the computation of analysis products. This STAR paper brings together researchers, developers and practitioners using in situ methods in extreme‐scale HPC with the goal to present existing methods, infrastructures, and a range of computational science and engineering applications using in situ analysis and visualization.

Current and upcoming cosmological observations allow us to probe structures on smaller and smaller scales, entering highly nonlinear regimes. In order to obtain theoretical predictions in these regimes, large cosmological simulations have to be carried out. The promised high accuracy from observations makes the simulation task very demanding: the simulations have to be at least as accurate as the observations. This requirement can only be fulfilled by carrying out an extensive code verification program. The first step of such a program is the comparison of different cosmology codes including gravitational interactions only. In this paper, we extend a recently carried out code comparison project to include five more simulation codes. We restrict our analysis to a small cosmological volume which allows us to investigate properties of halos. For the matter power spectrum and the mass function, the previous results hold, with the codes agreeing at the 10% level over wide dynamic ranges. We extend our analysis to the comparison of halo profiles and investigate the halo count as a function of local density. We introduce and discuss ParaView as a flexible analysis tool for cosmological simulations, the use of which immensely simplifies the code comparison task.

We present an architectural approach based on parallel data streaming to enable visualizations on a parallel cluster. Our approach requires less memory than other visualizations while achieving high code reuse. We implemented our architecture within the Visualization Toolkit (VTK). It includes specific additions to support message passing interfaces (MPIs); memory limit-based streaming of both implicit and explicit topologies; translation of streaming requests between topologies; and passing data and pipeline control between shared, distributed, and mixed memory configurations. The architecture directly supports both sort-first and sort-last parallel rendering.