TY - GEN
T1 - Exploring the effect of noise on the performance benefit of nonblocking allreduce
AU - Widener, Patrick
AU - Ferreira, Kurt B.
AU - Levy, Scott
AU - Hoefler, Torsten
N1 - Publisher Copyright:
© ACM 2014.
PY - 2014/9/9
Y1 - 2014/9/9
N2 - Relaxed synchronization offers the potential of maintaining application scalability by allowing many processes to make independent progress when some processes suffer delays. Yet, the benefits of this approach in important parallel workloads have not been investigated in detail. In this paper, we use a validated simulation approach to explore the noise mitigation effects of nonblocking allreduce in workloads where allreduce is a major contributor to total execution time. Although a nonblocking allreduce is unlikely to provide significant benefit to applications in the low-OS-noise environments expected in next-generation HPC systems, we show that it can potentially improve application runtime with respect to other noise types.
AB - Relaxed synchronization offers the potential of maintaining application scalability by allowing many processes to make independent progress when some processes suffer delays. Yet, the benefits of this approach in important parallel workloads have not been investigated in detail. In this paper, we use a validated simulation approach to explore the noise mitigation effects of nonblocking allreduce in workloads where allreduce is a major contributor to total execution time. Although a nonblocking allreduce is unlikely to provide significant benefit to applications in the low-OS-noise environments expected in next-generation HPC systems, we show that it can potentially improve application runtime with respect to other noise types.
KW - Collective operations
KW - Nonblocking collectives
KW - OS noise
UR - http://www.scopus.com/inward/record.url?scp=84958954568&partnerID=8YFLogxK
U2 - 10.1145/2642769.2642786
DO - 10.1145/2642769.2642786
M3 - Conference contribution
AN - SCOPUS:84958954568
T3 - ACM International Conference Proceeding Series
SP - 77
EP - 82
BT - Proceedings of the 21st European MPI Users' Group Meeting, EuroMPI/ASIA 2014
PB - Association for Computing Machinery
T2 - 21st European MPI Users' Group Meeting, EuroMPI/ASIA 2014
Y2 - 9 September 2014 through 12 September 2014
ER -