Banca de QUALIFICAÇÃO: CARLA DOS SANTOS SANTANA

Uma banca de QUALIFICAÇÃO de DOUTORADO foi cadastrada pelo programa.
STUDENT : CARLA DOS SANTOS SANTANA
DATE: 16/12/2021
TIME: 14:00
LOCAL: meet.google.com/urc-dbqd-fik
TITLE:

Fault Tolerance techniques applied to geophysical numerical methods


KEY WORDS:

fault tolerance, fault detection, replication, checkpointing, heartbeat message, watchdog, high-performance computing, full-waveform inversion


PAGES: 30
BIG AREA: Ciências Exatas e da Terra
AREA: Ciência da Computação
SUMMARY:

High-performance computing (HPC) has increased and provided the study of problems that involve several calculations and a significant amount of data (as geophysical methods) in a viable runtime.   One of the primary purposes of a problem that applies HPC techniques is scalability.  In other words, the application should maintain the same performance when the number of nodes grows. Scalability is also a problem because each node provides a particular mean time between failures  (MTBF)  therefore,  the more nodes are used,  the more elevated are the probabilities of failure.   An application that requires significant computation to be resilient is an essential feature,  and dealing with faults is critical to running it in some high-performance computing environments (HERAULT; ROBERT, 2015). HPC has been employed in geophysical methods for algorithms with high computational complexity, like Full waveform inversion (FWI). This method measures the seismic wave propagation velocity model from the difference between observed and artificially modeled data (VIRIEUX; OPERTO, 2009).   A failure in a subset of those nodes may cause an unrecoverable failure in FWI, which can produce a meaningful financial impact as it may take several days or weeks to recompute the lost data. The FWI needs a fault tolerance (FT) technique to have the ability to continue running even in the event of faults (KHAITAN, 2016). Completely reliable systems are impossible because failures are inevitable.  However, we can reduce the failures' consequences by applying suitable techniques (WEBER, 2003).  This work researches FT techniques (as replication, checkpointing, heartbeat message, and watchdog) to apply 3D FWI for the HPC environment.


BANKING MEMBERS:
Presidente - 1673543 - SAMUEL XAVIER DE SOUZA
Externa ao Programa - 1434796 - IDALMIS MILIAN SARDINA MARTINS
Externo ao Programa - 3216921 - TIAGO TAVARES LEITE BARROS
Notícia cadastrada em: 03/12/2021 09:34
SIGAA | Superintendência de Tecnologia da Informação - (84) 3342 2210 | Copyright © 2006-2024 - UFRN - sigaa12-producao.info.ufrn.br.sigaa12-producao