Interactive Exploration of Large Data in Hybrid Visualization Environments

Abstract

With rising data sizes and growing complexity, the results of modern numerical simulations are increasingly dif- ficult to understand. Thus, using Virtual Reality methodology for an interactive analysis of such data gains more and more importance. However, interaction within virtual environments comes at the cost of real-time constraints, which are difficult to meet. Using a hybrid visualization environment consisting of a high-performance computing (HPC) system connected to a graphics workstation (or multiple rendering nodes) we propose a workload distribution which significantly increases interactivity during the data analysis process. Based on a novel model of the exploration process, we introduce an additional step into the conventional visualization pipeline before mapping the whole process onto system components. This incorporates the respective benefits of high-performance computing and GPU-based computation into a single visualization framework. Basically, by coupling an HPC-based extraction of a region-of-interest to GPU-based flow visualization, an interactive analysis of large datasets is made possible. Taking interactive particle tracing and volume rendering as examples, we show the applicability of our approach to an interactive exploration of datasets exceeding the memory limits of a single workstation.