DistributedCompATM

DistributedCompATM - G EOGRAPHICALLY D ISTRIBUTED C...

Info iconThis preview shows pages 1–2. Sign up to view the full content.

View Full Document Right Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: G EOGRAPHICALLY D ISTRIBUTED C OMPUTING : ATM over the NASA ACTS Satellite P ATRICK W. D OWD 1 , F RANK A. P ELLEGRINO , 1 T ODD M. C ARROZZI , 1 , S ARAGUR M. S RINIDHI 2 1 Dept. Electrical & Computer Engineering 2 Sterling Software/NASA Lewis Research Center State University of New York at Buffalo 21000 Brookpark Road Buffalo, NY 14260 Cleveland, Ohio 44135 Abstract This paper outlines some of the problems and the solutionsdeveloped to support geographically distributed com- puting via ATM. In particular, applications developed with the Parallel Virtual Machine (PVM) [1] message passing library, communicating via ATM at OC3c speeds (155 Mbps) through the NASA ACTS satellite are considered. A primary goal of this work is to assess the suitability of an ATM-based network to support interprocess communication and remote file I/O sys- tems for distributed computing. This paper restricts itself to the behavior of PVM in a large propagation delay environment. Refer to [24] for additional detail and performance results. 1 Introduction There is interest in developing a capability of supporting ge- ographically distributed computing. This would allow more effective resource sharing and improved utilization of comput- ing resources. However, the propagation delay between two computer systems can severely hamper the achievable perfor- mance. In this paper we consider this problem, define three possible solutions to it and then illustrate the results of the solutions as implemented. We are primarily concerned with geographically dis- tributed computing, and use the NASA ACTS satellite to inves- tigate this problem. This poses a particularly extreme situation since the propagation delay is on the order of hundreds of mil- liseconds. The experimental scenario is illustrated in Figure 1, where the end points are a CRAY YMP supercomputer that is used for CFD computation, and an SGI Onyx which is used for visualization. A major issue to be examined is the use of ATM-based local-area and wide-area networks in distributed computing. In particular, a primary goal of this work is to assess the suitability This work was supported by NASA through Grant NAG3-1548. The authors can be reached through dowd,fap,carrozzi,saragur @lace.lerc.nasa.gov and would like to thank Dave Brooks of NASA/Lewis for his assistance. Preprints of this paper, and other related works, can be obtained through WWW at http://piranha.eng.buffalo.edu/ . ONYX ONYX SGI SGI SPARC 10 SUN SPARC 10 SUN Ethernet OC3c ATM OC3c HIPPI NASA ACTS Satelite BBG Fore ASX-200 OC3c- ATM- Fibre Channel- FDDI- ALLNODE- Ethernet Communication: Intra-Cluster IBM RS/6000 Inter-Cluster- Ethernet- ATM Comm CRAY YMP IBM SP/2 Figure 1: Experimental Scenario of an ATM-based network to support the interprocess commu- nication and remote file I/O system requirement of distributed computing....
View Full Document

Page1 / 5

DistributedCompATM - G EOGRAPHICALLY D ISTRIBUTED C...

This preview shows document pages 1 - 2. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online