oa Visualization Methods And Computational Steering Of The Electron Avalanches In The High-Energy Particle Detector Simulators

The traditional cycle in the simulation of the electron avalanches and any scientific simulation is to prepare input, execute a simulation, and to visualize the results as a post-processing step. Usually, such simulations are long running and computationally intensive. It is not unusual for a simulation to keep running for several days or even weeks. If the experiment leads to the conclusion that there is incorrect logic in the application, or input parameters were wrong, then simulation has to be restarted with correct parameters. A most common method of analyzing the simulation results is to gather the data on disk and visualize after the simulation finishes. The electron avalanche simulations can generate millions of particles that can require huge amount of disk I/O. The disk being inherently slow can become the bottleneck and can degrade the overall performance. Furthermore, these simulations are commonly run on the supercomputer. The supercomputer maintains a queue of researchers' programs and executes them as time and priorities permit. If the simulation produces incorrect results and there is a need to restart it with different input parameters, it may not be possible to restart it immediately because supercomputer is typically shared by several other researchers. The simulations (or jobs) have to wait in the queue until they are given a chance to execute again. It increases the scientific simulation cycle time and hence reduces the researcher's productivity. This research work proposes a framework to let researchers visualize the progress of their experiments so they could detect the potential errors at early stages. It will not only enhance their productivity but will also increase efficiency of the computational resources. This work focuses on the simulations of the propagation and interactions of electrons with ions in particle detectors known as Gas Electron Multipliers (GEMs). However, the proposed method is applicable to any scientific simulation from small to very large scale.