Interactive Hybrid Remote Rendering for Multi-pipe Powerwall Systems

Abstract

To understand complex physical phenomena, interactive and explorative post-processing of their numerical simulations has proven to be a feasible solution. This task can be enhanced by using virtual environments (VR) with high immersion and natural interaction techniques. While interactive VR systems provide high frame rates with low latencies, the hardware resources driving their displays are often insuffcient to visualize large scale simulation results. Local memory capacities are easily exceeded by dataset sizes and render performance is insuffcient to achieve the required frame rates. Parallel remote rendering techniques, on the other hand, support huge distributed memories and massive render performance. However these techniques have long interaction delays and the update rate is bandwidth limited. In this paper, we introduce a software architecture that oers a hybrid rendering approach for interactive visualization of large scale simulation data using a multi-pipe powerwall system. Local and remote images are generated on a frontend and parallel backend application with different frame rates. An image re-adjustment technique is used to deal with delayed remote images and hide latencies.