Messaging on gigabit ethernet: some experiments with GAMMA and other systems

作者: G. Ciaccio

DOI: 10.1109/IPDPS.2001.925148

关键词:

摘要: The Genoa Active Message MAchine (GAMMA) is a lightweight communication system based on the Ports paradigm, originally designed for efficient implementation over low-cost Fast Ethernet interconnects. In this paper we report about recently completed porting of GAMMA to Packet Engines GNIC-II and Netgear GA620 Gigabit adapters, provide comparison among GAMMA, MPI/GAMMA, TCP/IP, MPICH, such commodity interconnects, using different performance metrics. With combination low end-to-end latency (9.5 μs with GNIC-II, 32 GA620) high transmission throughput (almost 97 MByte/s 125 GA620, latter obtained without changing firmware adapter), demonstrates potential protocols yield messaging comparable best Myrinet-based systems. This result interest, given envisaged drop in cost due transition from fiber optic UTP cabling ever increasing mass market production standard interconnect. We also reports technique message fragmentation that commonly exploited increase short message. When different, though more widely used, metrics considered, results into loss rather than improvement.

参考文章(13)
Andrew Gallatin, Ken Yocum, Jeff Chase, Trapeze/IP: TCP/IP at near-gigabit speeds usenix annual technical conference. pp. 34- 34 ,(1999)
P. D. V. van der Stok, M. Boosten, R. W. Dobinson, MESH: MEssageing and ScHeduling for Fine-Grain Parallel Processing on Commodity Plattforms. parallel and distributed processing techniques and applications. pp. 1716- 1722 ,(1999)
Hermann Hellwagner, Markus Lachowitz, Matthias Ohlenroth, Exploring the Performance of VI Architecture Communication Features in the Giganet Cluster LAN. parallel and distributed processing techniques and applications. ,(2000)
Loic Prylli, Bernard Tourancheau, BIP: A new protocol designed for high performance networking on Myrinet international parallel processing symposium. pp. 472- 485 ,(1998) , 10.1007/3-540-64359-1_721
Hiroshi Tezuka, Atsushi Hori, Yutaka Ishikawa, Mitsuhisa Sato, PM: An Operating System Coordinated High Performance Communication Library ieee international conference on high performance computing data and analytics. pp. 708- 717 ,(1997) , 10.1007/BFB0031642
Mario Lauria, Scott Pakin, Andrew Chien, Efficient layering for high speed communication: the MPI over Fast Messages (FM) experience Cluster Computing. ,vol. 2, pp. 107- 116 ,(1999) , 10.1023/A:1019018423211
Giuseppe Ciaccio, Giovanni Chiola, GAMMA: A low-cost network of workstations based on active messages. PDP. pp. 78- 83 ,(1997)
Piyush Shivam, Pete Wyckoff, Dhabaleswar Panda, EMP Proceedings of the 2001 ACM/IEEE conference on Supercomputing (CDROM) - Supercomputing '01. pp. 57- 57 ,(2001) , 10.1145/582034.582091
Shinji Sumimoto, Hiroshi Tezuka, Atsushi Hori, Hiroshi Harada, Toshiyuki Takahashi, Yutaka Ishikawa, The design and evaluation of high performance communication using a Gigabit Ethernet international conference on supercomputing. pp. 260- 267 ,(1999) , 10.1145/305138.305202
G. Chiola, G. Ciaccio, Efficient parallel processing on low-cost clusters with GAMMA active ports parallel computing. ,vol. 26, pp. 333- 354 ,(2000) , 10.1016/S0167-8191(99)00108-8