TurboVNC Good way to provide access to remote visualization applications SARA Remote Visualization Serviceģ Large-scale visualization Large-scale parallel computations usually produce large-scale output data Gigabytes to terabytes (or more) data In general also need parallel processing and rendering in order to visualize Fast access to data storage necessary Approach A: visualize on HPC system where data was produced Problem #1 - HPC systems usually batch-oriented Creating visualizations has important interactive component Queue: need to wait for job to start, unpredictable when it runs Problem #2 - GPUs often not available in HPC systems Software-only 2D/3D rendering possible, but on the order of 5x - 50x slower than using a GPUĤ Large-scale visualization (2) Approach B: visualize locally I.e. 1 Remote visualization VirtualGL Paul Melis SARA Visualization GroupĢ Overview Remote visualization Why? What? How? VirtualGL Popular package for remote visualization Background, usage, features.
0 kommentar(er)
