Modeling the backoff strategies of access to shared-memory multiprocessors

被引:0
作者
Boguslavskii, LB [1 ]
Lyakhov, AI [1 ]
Sevcik, KC [1 ]
机构
[1] UNIV TORONTO,TORONTO,ON,CANADA
关键词
D O I
暂无
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
Consideration is given to a multiprocessor hierarchical architecture where the stations constituted by processor modules interconnected by a common bus are in turn connected by a ring-shaped system. To access a remote memory module, the processor passes a request via the interconnection system. If the required memory module is busy, then the request is rejected and the time of communication system is wasted. To reduce the time losses in the common bus of a multiprocessor station, the present publication proposes and models the backoff strategies for control of access to the shared memory. Using queueing theory and equilibrium analysis, we develop an approximate model of a generalized backoff strategy and a method for its analysis, which an based on realistic assumptions and enable one to evaluate the efficiency of the generalized strategy with low computational effort.
引用
收藏
页码:601 / 610
页数:10
相关论文
共 11 条
[1]  
CHAIKEN D, 1991, P 4 INT C ARCH SUPP, P224
[2]  
Fukuda A., 1986, Systems and Computers in Japan, V17, P84, DOI 10.1002/scj.4690170411
[3]   PERFORMANCE ANALYSIS OF FUTURE SHARED STORAGE-SYSTEMS [J].
GOYAL, A ;
AGERWALA, T .
IBM JOURNAL OF RESEARCH AND DEVELOPMENT, 1984, 28 (01) :95-108
[4]  
HOLLIDAY M, 1992, CS199218 DUKE U DEP, P27708
[5]  
HOOGENDOORN CM, 1977, IEEE T COMPUT, V26, P990
[6]   THE STANFORD DASH MULTIPROCESSOR [J].
LENOSKI, D ;
LAUDON, J ;
GHARACHORLOO, K ;
WEBER, WD ;
GUPTA, A ;
HENNESSY, J ;
HOROWITZ, M ;
LAM, MS .
COMPUTER, 1992, 25 (03) :63-79
[7]  
PATEL JH, 1982, IEEE T COMPUT, V31, P296, DOI 10.1109/TC.1982.1675995
[8]   PERFORMANCE BENEFITS AND LIMITATIONS OF LARGE NUMA MULTIPROCESSORS [J].
SEVCIK, KC ;
ZHOU, SN .
PERFORMANCE EVALUATION, 1994, 20 (1-3) :185-205
[9]  
SHIN KG, 1982, IEEE T COMPUT, V31, P1045, DOI 10.1109/TC.1982.1675921
[10]  
VRANESIC ZG, 1991, IEEE COMPUT, V24, P72