TY - GEN
T1 - An analysis of system balance requirements for scientific applications
AU - Alam, Sadaf R.
AU - Vetter, Jeffrey S.
PY - 2006
Y1 - 2006
N2 - Scientific applications are diverse in terms of the resource requirements, and tend to vary significantly from commercial applications. In order to provide sustained performance, a target high performance computing (HPC) platform must offer a balance between CPU performance to memory, interconnect and I/O subsystems performance. We characterize the system balance requirements for two largescale Office of Science applications, GYRO (fusion simulation) and POP (climate modeling), and develop platform-independent parameterized requirement models. We measure the parallel efficiencies for GYRO and POP on three multiprocessor systems: an SMP cluster (IBM p690), a shared-memory system (SGIAltix) and a vector supercomputer (Cray X1). The higher computational intensity and interconnect bandwidth requirements of GYRO result in higher performance efficiencies on the vector platform. At the same time, small message sizes in POP benefit from low MPI latencies of the shared-memory platform. Overall results confirm system balance requirements that are generated by the requirement models.
AB - Scientific applications are diverse in terms of the resource requirements, and tend to vary significantly from commercial applications. In order to provide sustained performance, a target high performance computing (HPC) platform must offer a balance between CPU performance to memory, interconnect and I/O subsystems performance. We characterize the system balance requirements for two largescale Office of Science applications, GYRO (fusion simulation) and POP (climate modeling), and develop platform-independent parameterized requirement models. We measure the parallel efficiencies for GYRO and POP on three multiprocessor systems: an SMP cluster (IBM p690), a shared-memory system (SGIAltix) and a vector supercomputer (Cray X1). The higher computational intensity and interconnect bandwidth requirements of GYRO result in higher performance efficiencies on the vector platform. At the same time, small message sizes in POP benefit from low MPI latencies of the shared-memory platform. Overall results confirm system balance requirements that are generated by the requirement models.
UR - http://www.scopus.com/inward/record.url?scp=34547420620&partnerID=8YFLogxK
U2 - 10.1109/ICPP.2006.21
DO - 10.1109/ICPP.2006.21
M3 - Conference contribution
AN - SCOPUS:34547420620
SN - 0769526365
SN - 9780769526362
T3 - Proceedings of the International Conference on Parallel Processing
SP - 229
EP - 236
BT - ICPP 2006
T2 - ICPP 2006: 2006 International Conference on Parallel Processing
Y2 - 14 August 2006 through 18 August 2006
ER -