Improving I/O Performance of RTM Algorithm for Oil and Gas Simulation

• Pablo J. Pavan UFRGS
• Matheus S. Serpa UFRGS
• Edson L. Padoin UNIJUI
• Lucas M. Schnorr UFRGS
• Philippe O. A. Navaux UFRGS
• Jairo Panetta ITA

Resumo

Computational simulations of physical phenomena are only possible due to the computational resources available in High Performance Computing (HPC) systems. These applications generally generate a large volume of data which is read and written to disk during the simulation. Since these operations are costly, this becomes one of the performance bottlenecks of these applications. In this way, this work analyzes and proposes optimizations for the input and output operations of a geophysical application that uses the Reverse Time Migration (RTM) algorithm. The results show that using checkpoints and increasing the size of the request reduces application execution time by up to 17.33%.

Palavras-chave: Computational modeling, Mathematical model, High performance computing, Geophysics, Oils, Computerscience, Informatics, Oil and Gas Simulation, parallel I/O, storage, high performance computing