The use of parallel computing makes it feasible to simulate realistic seismological events, whose reconstruction requires wide domains, high frequencies and the introduc- tion of the dissipation terms. The propagation problem of seismic waves is a key feature of the earthquake dynamics that we are interested in numerically modeling and simu- lating. In particular, in this work we present several preliminary results about the load balancing for the parallel resolution of a simulation of the propagation of seismic waves in a 3D heterogeneous medium. The Finite Element Method is employed for the spatial discretization by using non–structured tetrahedral meshes. The Newmark method is used for the time discretization. With the aim to study a priori the load balancing, we intro- duce two performance indices: closing nodes and node balancing. In particular, the first one estimates the amount of processor data, the latter provides information on work–load distribution. The variation of these indices as functions of the number of processors and of the number of nodes of the grid is then investigated.
Simulation of seismic wave propagation in 3D heterogeneous media:a parallel computing approach
AGRESTE, SANTA;RICCIARDELLO, ANGELA
2011-01-01
Abstract
The use of parallel computing makes it feasible to simulate realistic seismological events, whose reconstruction requires wide domains, high frequencies and the introduc- tion of the dissipation terms. The propagation problem of seismic waves is a key feature of the earthquake dynamics that we are interested in numerically modeling and simu- lating. In particular, in this work we present several preliminary results about the load balancing for the parallel resolution of a simulation of the propagation of seismic waves in a 3D heterogeneous medium. The Finite Element Method is employed for the spatial discretization by using non–structured tetrahedral meshes. The Newmark method is used for the time discretization. With the aim to study a priori the load balancing, we intro- duce two performance indices: closing nodes and node balancing. In particular, the first one estimates the amount of processor data, the latter provides information on work–load distribution. The variation of these indices as functions of the number of processors and of the number of nodes of the grid is then investigated.Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.