Shared Memory Model over a Switchless PCIe NTB Interconnect Network
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
The role of the interconnect network, which connects computing nodes to each other, is important in highperformancecomputing (HPC) systems. In recent years, the peripheral component interconnect express (PCIe)has become a promising interface as an interconnection network for high-performance and cost-effective HPCsystems having the features of non-transparent bridge (NTB) technologies. OpenSHMEM is a programmingmodel for distributed shared memory that supports a partitioned global address space (PGAS). Currently, littlework has been done to develop the OpenSHMEM library for PCIe-interconnected HPC systems. This paperintroduces a prototype implementation of the OpenSHMEM library through a switchless interconnect networkusing PCIe NTB to provide a PGAS programming model. In particular, multi-interrupt, multi-thread-based datatransfer over the OpenSHMEM shared memory model is applied at the implementation level to reduce thelatency and increase the throughput of the switchless ring network system. The implemented OpenSHMEMprogramming model over the PCIe NTB switchless interconnection network provides a feasible, cost-effectiveHPC system with a PGAS programming model.
Abstract
The role of the interconnect network, which connects computing nodes to each other, is important in highperformancecomputing (HPC) systems. In recent years, the peripheral component interconnect express (PCIe)has become a promising interface as an interconnection network for high-performance and cost-effective HPCsystems having the features of non-transparent bridge (NTB) technologies. OpenSHMEM is a programmingmodel for distributed shared memory that supports a partitioned global address space (PGAS). Currently, littlework has been done to develop the OpenSHMEM library for PCIe-interconnected HPC systems. This paperintroduces a prototype implementation of the OpenSHMEM library through a switchless interconnect networkusing PCIe NTB to provide a PGAS programming model. In particular, multi-interrupt, multi-thread-based datatransfer over the OpenSHMEM shared memory model is applied at the implementation level to reduce thelatency and increase the throughput of the switchless ring network system. The implemented OpenSHMEMprogramming model over the PCIe NTB switchless interconnection network provides a feasible, cost-effectiveHPC system with a PGAS programming model.
참고문헌 (23)
- "Interconnect family statistics" (2018)[URL]
- "Peripheral Component Interconnect Special Interest Group"[URL]
- 김영우, "PCI Express 기반 시스템 인터커넥트의 설계 및 구현", 전자공학회논문지, 52(8), pp.74~86 (2015), 대한전자공학회
- L. Mohrmann, "Creating multicomputer test systems using PCI and PCI Express", Proceedings of 2009 IEEE International Automatic Testing Conference (AUTOTESTCON), pp.7~10 (2009)
- A. Richter, "Resolving performance interference in SR-IOV setups with PCIe Quality-of-Service extensions", Proceedings of 2016 Euromicro Conference on Digital System Design (DSD), pp.454~462 (2016)
- J. Liu, "Evaluating infiniband performance with PCI express", IEEE Micro, 25(1), pp.20~29 (2005)
- H. Wong, "PCI express multi-root switch reconfiguration during system operation", Ph.D., pp.2011
- V. Krishnan, "Towards an integrated IO and clustering solution using PCI express", Proceedings of 2007IEEE International Conference on Cluster Computing, pp.259~266 (2007)
- "Non-transparent Bridging with IDT 89HPES32NT24G2 PCI Express NTB Switch" (2019)[URL]
- W. C. C. Tu, "Seamless fail-over for PCIe switched networks", Proceedings of the 11th ACM International Systems and Storage Conference, pp.101~111 (2018)
- "An introduction to the Partitioned Global Address Space (PGAS) programming model" (2009)[URL]
- B. Chapman, "Introducing OpenSHMEM:SHMEM for the PGAS community", Proceedings of the 4th Conference on Partitioned Global Address Space Programming Model, pp.1~3 (2010)
- J. R. Hammond, "OpenSHMEM and Related Technologies", pp.44~58 (2014), Springer
- S. Pophale, "OpenSHMEM performance and potential : a NPB experimental study", Proceedings of the 6th Conference on Partitioned Global Address Space Programming Models (PGAS) (2012)
- R. E. Kessler, "CRAY T3D: a new dimension for Cray Research", Proceedings of COMPCON Spring Digest of Papers, pp.176~182 (1993)
- "PEX 8749" (2011)[URL]
- 최민, "Feasibility and Performance Analysis of RDMA Transfer through PCI Express", JIPS(Journal of Information Processing Systems), 13(1), pp.95~103 (2017), 한국정보처리학회[DOI]
- L. Rota, "A new DMA PCIe architecture for Gigabyte data transmission", Proceedings of 2014 19th IEEE-NPSS Real Time Conference, pp.1~2 (2014)
- S. H. Lim, "Developing an OpenSHMEM model over a switchless PCIe nontransparent bridge interface", Proceedings of 2019 IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW), pp.593~602 (2019)
- "PCIe Networked Flash Storage" (2018)[URL]
- "Intel Xeon Processor C5500/C3500 Series Non-Transparent Bridge" (2010)[URL]
- J. M. Mellor-Crummey, "Algorithms for scalable synchronization on shared-memory multiprocessors", ACM Transactions on Computer Systems, 9(1), pp.21~65 (1991)
- C. Shim, "Design and implementation of initial OpenSHMEM on PCIe NTB based cloud computing", Cluster Computing, 22(1), pp.1815~1826 (2019)