The Impact on Latency and Bandwidth for a Distributed Shared Memory System Using a Gigabit Network Supporting the Virtual Interface Architecture

Previous studies have shown significant performance advantages in using Virtual Interface Architecture (VIA) instead of TCP/IP for handling network communication in the structured distributed shared memory system, PastSet. With the availability of network hardware that supports VIA, we wish to examine whether, and to what extend, an available hardware supported VIA implementation outperforms the softwareonly implementation for PastSet DSM. To do this, PastSet has been ported to two VIA implementations: M-VIA, which is a software implementation that we use on a 100 Mbit Fast Ethernet, and Giganet cLAN, which uses dedicated VIA hardware. The two implementations are tested, and performance results are compared with the reference TCP/IP implementation on the 100

[1]  Ewing Lusk,et al.  User''s Guide to the p4 Parallel Programming System , 1992 .

[2]  Kai Li,et al.  Retrospective: virtual memory mapped network interface for the SHRIMP multicomputer , 1994, ISCA '98.

[3]  Peter Druschel,et al.  Operating systems and network interfaces , 1998 .

[4]  Edward W. Felten,et al.  Reducing waiting costs in user-level communication , 1997, Proceedings 11th International Parallel Processing Symposium.

[5]  Michael D. Smith,et al.  The measured performance of personal computer operating systems , 1996, TOCS.

[6]  Ami Marowka,et al.  The GRID: Blueprint for a New Computing Infrastructure , 2000, Parallel Distributed Comput. Pract..

[7]  Henri E. Bal,et al.  Efficient multicast on Myrinet using link-level flow control , 1998, Proceedings. 1998 International Conference on Parallel Processing (Cat. No.98EX205).

[8]  Miss A.O. Penney.  (b) , 1974, The New Yale Book of Quotations.

[9]  Liviu Iftode,et al.  Software Distributed Shared Memory over Virtual Interface Architecture: Implemenation and Performance , 2000, Annual Linux Showcase & Conference.