TY - GEN
T1 - Scalable memory use in MPI
T2 - 18th European Message Passing Interface Users' Group Meeting, EuroMPI 2011
AU - Goodell, David
AU - Gropp, William
AU - Zhao, Xin
AU - Thakur, Rajeev
N1 - Funding Information:
Acknowledgments. This work was supported by the U.S. Department of Energy, under Contract DE-AC02-06CH11357 and Award DE-FG02-08ER25835, and by the National Science Foundation Grant #0702182.
PY - 2011
Y1 - 2011
N2 - One of the factors that can limit the scalability of MPI to exascale is the amount of memory consumed by the MPI implementation. In fact, some researchers believe that existing MPI implementations, if used unchanged, will themselves consume a large fraction of the available system memory at exascale. To investigate and address this issue, we undertook a study of the memory consumed by the MPICH2 implementation of MPI, with a focus on identifying parts of the code where the memory consumed per process scales linearly with the total number of processes. We report on the findings of this study and discuss ways to avoid the linear growth in memory consumption. We also describe specific optimizations that we implemented in MPICH2 to avoid this linear growth and present experimental results demonstrating the memory savings achieved and the impact on performance.
AB - One of the factors that can limit the scalability of MPI to exascale is the amount of memory consumed by the MPI implementation. In fact, some researchers believe that existing MPI implementations, if used unchanged, will themselves consume a large fraction of the available system memory at exascale. To investigate and address this issue, we undertook a study of the memory consumed by the MPICH2 implementation of MPI, with a focus on identifying parts of the code where the memory consumed per process scales linearly with the total number of processes. We report on the findings of this study and discuss ways to avoid the linear growth in memory consumption. We also describe specific optimizations that we implemented in MPICH2 to avoid this linear growth and present experimental results demonstrating the memory savings achieved and the impact on performance.
UR - http://www.scopus.com/inward/record.url?scp=80052990364&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=80052990364&partnerID=8YFLogxK
U2 - 10.1007/978-3-642-24449-0_17
DO - 10.1007/978-3-642-24449-0_17
M3 - Conference contribution
AN - SCOPUS:80052990364
SN - 9783642244483
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 140
EP - 149
BT - Recent Advances in the Message Passing Interface - 18th European MPI Users' Group Meeting, EuroMPI 2011, Proceedings
Y2 - 18 September 2011 through 21 September 2011
ER -