Workload scheduling analysis in geophysical problems
Workload scheduling, seismic processing, parallel processing.
This study addresses the Workload Scheduling analysis in two geophysical problems, RTM (Reverse Time Migration) and FWI (Full Waveform Inversion). The two techniques are computationally intensive with high use of the RAM memory and long execution time. A way of improving these methods is by using parallel computing. For the implementation in parallel environments, it is important to define the tasks (processes) distribution among the computational nodes.
This work presents the article "Workload scheduling comparison in a full waveform inversion distributed memory implementation" accepted for presentation in the 16th International Conference of the Brazilian Geophysical Society (SBGf), in which we compare two workload scheduling approaches for a parallel 2D acoustic FWI implementation: the centralized dynamic and the decentralized static. It also presents the next steps for the master research which include a centralized dynamic scheduling for RTM and another scheduling with load balancing for FWI 2D: work stealing. Finally, the following research steps are detailed and illustrated in a time schedule.