DocumentCode
2964496
Title
Multiple bus-based hierarchical multiprocessors and their bandwidth analysis
Author
Mahmud, Sycd Masud ; Samaratunga, L. Fissa ; Munteanu, Mircea D.
Author_Institution
Dept. of Electr. & Comput. Eng., Wayne State Univ., Detroit, MI, USA
fYear
1996
fDate
11-13 Jun 1996
Firstpage
311
Lastpage
318
Abstract
Recently a number of researchers have paid attention to the design and analysis of cluster-based multiprocessor systems. The design of cluster based systems is very appealing due to the fact that such systems require very inexpensive interconnection network compared to non-cluster based systems, especially for a large number of processors and memory modules. The design and analysis of a generalized hierarchical multiprocessor system was described by I.O. Mahgoub and A.K. Elmagarmid (1992). This multiprocessor system uses a hierarchical interconnection network, which is built using a number of small crossbar switches. In this paper we propose a different variation of the system presented previously. Our proposed system can be built by replacing every crossbar switch by a number of parallel buses. Since parallel buses provide better fault tolerance than a crossbar switch, the system proposed, in this paper is fault tolerant. Simulation models have been developed to verify the accuracy of the analytical models. The results obtained from the analytical models match closely to those obtained from the simulation models
Keywords
digital simulation; fault tolerant computing; multiprocessor interconnection networks; bandwidth analysis; cluster-based multiprocessor systems; fault tolerance; generalized hierarchical multiprocessor system; multiple bus-based hierarchical multiprocessors; parallel buses; simulation models; Analytical models; Bandwidth; Communication switching; Fault tolerance; Fault tolerant systems; Hierarchical systems; Multiprocessing systems; Multiprocessor interconnection networks; Performance analysis; Switches;
fLanguage
English
Publisher
ieee
Conference_Titel
Algorithms & Architectures for Parallel Processing, 1996. ICAPP 96. 1996 IEEE Second International Conference on
Print_ISBN
0-7803-3529-5
Type
conf
DOI
10.1109/ICAPP.1996.562890
Filename
562890
Link To Document