Parallel Caching for Scientific Visualization

Moving the simulation results produced by thousands of computing cores to the scientist office is not anymore an option. Remote visualization proposes to perform heavy duty postprocessing tasks at the computing center while transferring images to the scientist. To be effective, such environement needs to be flexible and interactive. Data loading from disks, processing, transfers to the client on long distance networks can incur a significant latency that impairs interactivity. In this paper we propose to cache the data at various levels along the processing pipeline to overlap the processing and data transfer times when analyzing time varying data sets. Results show that latency can be divided by about 4 using more than 2 parallel servers.