TCP/IP versus VIA on Network of Workstations
Resumo
This paper evaluates the performance gains provided by VIA, ao user-level communication protocol, when compared to TCP/IP, a traditional, multilayered communication protocol. To achieve this purpose we run five distinct applications using the same network card and switch, and just change the communication protocol. For all but one application, the speedup of TCP/IP was between 28% and 97% that of achieved by VIA. We further run the same applications with a hardware implementation of VIA, to evaluate the gains that a hardware implementation could offer over a software implementation. For applications with high bandwidth demand, the hardware support helped to improve performance from 20% to 114%. Surprisingly, for one application, TCP/IP performs equally to the hardware implementation.
Referências
BRAUN, T.; DIOT, C. Protocol implementation using integrated layer processing. ln ACM SIGCOMM '95, Cambridge, MA, USA, August. 1995.
KAY, J.; PASQUALE, J. The lmportance of Non-Data Touching Processing Overheads in TCP/IP. Computer Communication Review, vol. 23, (no. 4), Oct. 1993, pp. 259-268.
EICKEN, T. von; CULLER, D.; GOLDSTEIN, S.; SCHAUSER, K. Active messages: a Mechanism for Integrated Communication and Computation. In Proceedings of the 19th Annual International Symposium on Computer Architecture, Gold Coast, Australia, May 1992, pp.256-266.
PAKIN, S.; KARAMCHETI, V.; CHIEN, A. Fast messages: efficient, portable communication for workstation clusters and MPPs. IEEE Concurrency, vol.5, (no.2), April-June 1997, pp.60-72.
EICKEN, T. von; BASU, A.; BUCH, V.; VOGELS, W. U-Net: A User-level Network Interface of parallel and Distributed Computing. In Proc. of the 15th ACM Symposium of Operating Systems Principies, vol. 29, (no.5), December 1995, pp. 40-53.
BLUMRICH, M. et al. Virtual Memory Mapped Network Interface for the SHRIMP Multicomputer. In Proceedings of the 21st International Symposium on Computer Architecture, April 1994, pp. 142-153.
LU, H.; DVARKADAS, S.; COX, A.; ZWAENEPOEL, W. Message Passing Versus Distributed Shared Memory on Networks of Workstations. In Supercomputing '95, 1995.
Compaq, Intel, and Microsoft. Virtual Interface Architecture Specification, Version 1.0. Available at http://www.viarch.org.
Compaq, Dell, Hewlett-Packard, IBM, Intel, Microsoft and Sun Microsystems. InfiniBand Trade Association. http://www.infinibandta.org
SINGH, J.; WEBER, W.; GUPTA, A. SPLASH: Stanford Parallel Applications for Shared-Memory. Computer Architecture News, 20(1):2-12, March 1992.
BAILEY, D.; BARTON, J.; LASINSKI, T. ; SIMON, H. The NAS Parallel Benchmarks. Technical Report 103863, NASA, July 1993.
National Energy Research Scientific Computing Center. M-V!A: A High Performance Modular VIA for Linux. http://www.nersc.gov/research/FTG/vial.
National Energy Research Scientific Computing Center. M VICH: MPI for Virtual Interface Architecture. http://www.nersc.gov/research/FTG/mvich.
Giganet Inc. http://www.giganet.com/
Message Passing Interface Forum. http://www.mpiforum.org/.
GEIST, G.; SUNDERAM, V. Network based concurrent computing on the PVM system. Concurrency: Practice and Experience, June 1992, pp. 293-311.
