DocumentCode
2979210
Title
Cluster technologies for high performance computing
Author
Ishii, Mitsuo
Author_Institution
Comput. Syst. Group, Fujitsu Labs. Ltd., Japan
fYear
1999
fDate
1999
Firstpage
168
Lastpage
170
Abstract
A cluster system is extending its ability, owing to the increase of transmission speed and interface speed between nodes and interconnect. Synfinity interconnect can carry out the routing and transfer the massages at the speed of gigabyte/s. The bottleneck problem raised in a conventional cluster system must be solved by improvements such as transmission speed and protocol. Based on the standard interface, any kind of commodity like the latest PCs, WSs and SMPs can be applied as a node computer. That makes the cluster system more attractive from the point of view of availability, usability, flexibility, scalability and cost performance. Moreover, a parallel program on the cluster system can be written by using a newly-developed message communication library, without paying any attention to the communication between inter- and intra-node. The cluster system will be also used instead of a parallel computer, MPP and the so called supercomputer, and it has new features of shared memory view
Keywords
message passing; multiprocessor interconnection networks; parallel programming; workstation clusters; cluster system; cluster technologies; high performance computing; interface speed; message communication library; node computer; parallel program; shared memory view; standard interface; synfinity interconnect; transmission speed; Availability; Computer interfaces; Costs; High performance computing; Libraries; Personal communication networks; Protocols; Routing; Scalability; Usability;
fLanguage
English
Publisher
ieee
Conference_Titel
Parallel Architectures, Algorithms, and Networks, 1999. (I-SPAN '99) Proceedings. Fourth InternationalSymposium on
Conference_Location
Perth/Fremantle, WA
ISSN
1087-4089
Print_ISBN
0-7695-0231-8
Type
conf
DOI
10.1109/ISPAN.1999.778934
Filename
778934
Link To Document