In Situ and In Transit Visualization for Numerical Simulations in HPC
Event Type
Registration Categories
Data Analytics
Scientific Computing
TimeMonday, 18 November 201912pm - 12:05pm
DescriptionTo overcome the difficulty in visualizing large-scale simulation results in HPC, we implemented in situ and in transit visualization using VisIt/Libsim and ADIOS2 into an in-house application for computational fluid dynamics, and evaluated both forms on a heterogeneous HPC system of the Japan Aerospace Exploration Agency. It is found that both frameworks can be used for batch and interactive visualization commonly conducted by researchers and engineers. The in situ approach costs less to implement in applications and is easier to use. Conversely, the in transit approach is advantageous in terms of lower execution overhead and its flexibility for using separate resources in a heterogeneous HPC system. It is therefore desirable to use both approaches as needed. Based on the knowledge obtained herein, this paper summarizes suggestions for research and development of the in situ and in transit visualization frameworks, the HPC system, and queue policy for interactive jobs, in order to realize in situ and in transit visualization in a heterogeneous HPC system.
Back To Top Button