DocumentCode :
2847433
Title :
Exploiting Latent I/O Asynchrony in Petascale Science Applications
Author :
Widener, Patrick ; Payne, Mary ; Bridges, Patrick ; Wolf, Matthew ; Abbasi, Hasan ; McManus, Scott ; Schwan, Karsten
Author_Institution :
Dept. of Comput. Sci., Univ. of New Mexico Albuquerque, Albuquerque, NM, USA
fYear :
2009
fDate :
22-25 Sept. 2009
Firstpage :
105
Lastpage :
112
Abstract :
We present a collection of techniques for exploiting latent I/O asynchrony which can substantially improve performance in data-intensive parallel applications. Latent asynchrony refers to an application´s tolerance for decoupling ancillary operations from its core computation, and is a property of HPC codes not fully explored by current HPC I/O systems. Decoupling operations such as buffering and staging, reorganization, and format conversion in space and in time from core codes can shorten I/O phases, preserving valuable MPP compute cycles. We describe in this paper DataTaps, IOgraphs, and Metabots, three tools which allow HPC developers to implement decoupled I/O operations. Using these tools, asynchrony can be exploited by data generators which overlap computation with communication, and by data consumers that perform data conversion and reorganization out-of-band and on-demand. In the context of a data-intensive fusion simulation, we show that exploiting latent asynchrony through decoupling of operations can provide significant performance benefits.
Keywords :
input-output programs; parallel programming; DataTaps tool; HPC I/O systems; I/O decoupling operations; IOgraphs tool; Metabots tool; data-intensive parallel application; high performance computing; latent asynchrony; petascale science application; Application software; Bandwidth; Bridges; Computer science; Concurrent computing; Data conversion; Filtering; Monitoring; Parallel processing; Petascale computing;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Parallel Processing Workshops, 2009. ICPPW '09. International Conference on
Conference_Location :
Vienna
ISSN :
1530-2016
Print_ISBN :
978-1-4244-4923-1
Electronic_ISBN :
1530-2016
Type :
conf
DOI :
10.1109/ICPPW.2009.67
Filename :
5365163
Link To Document :
بازگشت