Architecture of the Component Collective Messaging Interface

Different programming paradigms utilize a variety of collective communication operations, often with different semantics. We present the component collective messaging interface (CCMI) that can support asynchronous non-blocking collectives and is extensible to different programming paradigms and architectures. CCMI is designed with components written in the C++ programming language, allowing it to be reusable and extendible. Collective algorithms are embodied in topological schedules and executors that execute them. Portability across architectures is enabled by the multisend data movement component. CCMI includes a programming language adaptor used to implement different APIs with different semantics for different paradigms. We study the effectiveness of CCMI on 16K nodes of Blue Gene/P machine and evaluate its performance for the barrier, broadcast, and allreduce collective operations and several application benchmarks. We also present the performance of the barrier collective on the Abe Infiniband cluster.

[1]  George L.-T. Chiu,et al.  Overview of the Blue Gene/L system architecture , 2005, IBM J. Res. Dev..

[2]  Philip Heidelberger,et al.  The deep computing messaging framework: generalized scalable message passing on the blue gene/P supercomputer , 2008, ICS '08.

[3]  Torsten Hoefler,et al.  Implementation and performance analysis of non-blocking collective operations for MPI , 2007, Proceedings of the 2007 ACM/IEEE Conference on Supercomputing (SC '07).

[4]  Wu-chun Feng,et al.  The Quadrics Network: High-Performance Clustering Technology , 2002, IEEE Micro.

[5]  Debra Hensgen,et al.  Two algorithms for barrier synchronization , 1988, International Journal of Parallel Programming.

[6]  Alex Rapaport,et al.  Mpi-2: extensions to the message-passing interface , 1997 .

[7]  Kevin T. Pedretti,et al.  SMARTMAP: Operating system support for efficient data sharing among processes on a multi-core processor , 2008, 2008 SC - International Conference for High Performance Computing, Networking, Storage and Analysis.

[8]  Xin Yuan,et al.  A study of process arrival patterns for MPI collective operations , 2007, ICS.

[9]  Forum Mpi.  MPI: A Message-Passing Interface , 1994 .

[10]  Keith Refson,et al.  Moldy: a portable molecular dynamics simulation program for serial and parallel computers , 2000 .

[11]  Steven J. Middleton,et al.  Model of very fast (> 75 Hz) network oscillations generated by electrical coupling between the proximal axons of cerebellar Purkinje cells , 2008, The European journal of neuroscience.

[12]  Rajeev Thakur,et al.  Optimization of Collective Communication Operations in MPICH , 2005, Int. J. High Perform. Comput. Appl..

[13]  William Gropp,et al.  Design and implementation of message-passing services for the Blue Gene/L supercomputer , 2005, IBM J. Res. Dev..

[14]  George Almási,et al.  Performance without pain = productivity: data layout and collective communication in UPC , 2008, PPoPP.

[15]  Philip Heidelberger,et al.  Optimization of All-to-All Communication on the Blue Gene/L Supercomputer , 2008, 2008 37th International Conference on Parallel Processing.

[16]  Xin Yuan,et al.  STAR-MPI: self tuned adaptive routines for MPI collective operations , 2006, ICS '06.

[17]  Philip Heidelberger,et al.  Optimization of MPI collective communication on BlueGene/L systems , 2005, ICS '05.

[18]  Ibm Blue,et al.  Overview of the IBM Blue Gene/P Project , 2008, IBM J. Res. Dev..

[19]  Wei Huang,et al.  Design of High Performance MVAPICH2: MPI2 over InfiniBand , 2006, Sixth IEEE International Symposium on Cluster Computing and the Grid (CCGRID'06).