Developing an OpenSHMEM Model Over a Switchless PCIe Non-Transparent Bridge Interface

OpenSHMEM is an emerging parallel programming model in distributed systems that supports partitioned global address space (PGAS). It was originally developed for Cray systems, and it was subsequently adopted by numerous vendors of the InfiniBand and Ethernet network interfaces, since most high-performance computing (HPC) systems deploy switch-based interconnect networks using InfiniBand and Ethernet technologies. Recently, Peripheral Component Interconnect Express (PCIe) has become one of the most promising candidates for deploying cost-effective popular HPC systems because of its low cost and powerful features, as well as the Non-Transparent Bridge(NTB) technology (interconnect interface for PCIe). However, there is little work done on implementing the OpenSHMEM library for PCIe systems. Herein, we introduce a prototype of a switchless interconnect network with PCIe NTB. In our switchless interconnect network system, the computing nodes are interconnected via the PCIe NTB interface. Based on the PCIe NTB interconnect network, we implemented the OpenSHMEM programming interface to support the PGAS mechanism for PCIe NTB interconnect networks. Our design and implementation of an OpenSHMEM programming model via PCIe NTB shows that is feasible and possible to have a PCIe-based, cost-effective, high performance interconnect network in a high-performance programming model.

[1]  Barbara M. Chapman,et al.  Implementing OpenSHMEM Using MPI-3 One-Sided Communication , 2014, OpenSHMEM.

[2]  W Wong PCI Express Multi-Root Switch Reconfiguration During System Operation by Heymian , 2011 .

[3]  A. Kopmann,et al.  A new DMA PCIe architecture for Gigabyte data transmission , 2014, 2014 19th IEEE-NPSS Real Time Conference.

[4]  Vivek Sarkar,et al.  Implementation and Evaluation of OpenSHMEM Contexts Using OFI Libfabric , 2017, OpenSHMEM.

[5]  Jeffery A Kuehn,et al.  OpenSHMEM Performance and Potential: A NPB Experimental Study , 2012 .

[6]  Michael L. Scott,et al.  Algorithms for scalable synchronization on shared-memory multiprocessors , 1991, TOCS.

[7]  Lee Mohrmann,et al.  Creating multicomputer test systems using PCI and PCI Express , 2009, 2009 IEEE AUTOTESTCON.

[8]  Min Choi,et al.  Design and implementation of initial OpenSHMEM on PCIe NTB based cloud computing , 2018, Cluster Computing.

[9]  Manjunath Gorentla Venkata,et al.  Designing a High Performance OpenSHMEM Implementation Using Universal Common Communication Substrate as a Communication Middleware , 2014, OpenSHMEM.

[10]  Barbara M. Chapman,et al.  Introducing OpenSHMEM: SHMEM for the PGAS community , 2010, PGAS '10.

[11]  Venkata Krishnan Towards an integrated IO and clustering solution using PCI express , 2007, 2007 IEEE International Conference on Cluster Computing.

[12]  J. Respondek Numerical approach to the non-linear diofantic equations with applications to the controllability of infinite dimensional dynamical systems , 2005 .

[13]  Young-Woo Kim,et al.  Design and Implementation of an Alternate System Interconnect based on PCI Express , 2015 .

[14]  Amith R. Mamidala,et al.  Evaluating InfiniBand performance with PCI Express , 2005, IEEE Micro.

[15]  Jong Hyuk Park,et al.  Feasibility and Performance Analysis of RDMA Transfer through PCI Express , 2017, J. Inf. Process. Syst..

[16]  Tzi-cker Chiueh,et al.  Seamless Fail-over for PCIe Switched Networks , 2018, SYSTOR.