The Approximate Agreement Problem of Massively Parallel Systems

博士 === 國立交通大學 === 資訊工程研究所 === 82 === An appropiate approach--neighborhood approach--is proposed forng an approximate agreement in massively prallel systems. Instead-to-all broadcasting that conventional approaches adopt, pproach requires only direct ne...

Full description

Bibliographic Details
Main Authors: Ren-Liang Cheng, 鄭仁亮
Other Authors: Chung-Ping Chung
Format: Others
Language:en_US
Published: 1994
Online Access:http://ndltd.ncl.edu.tw/handle/69764168854156157955
id ndltd-TW-082NCTU0392072
record_format oai_dc
spelling ndltd-TW-082NCTU03920722016-07-18T04:09:34Z http://ndltd.ncl.edu.tw/handle/69764168854156157955 The Approximate Agreement Problem of Massively Parallel Systems 大量平行系統之近似同意問題 Ren-Liang Cheng 鄭仁亮 博士 國立交通大學 資訊工程研究所 82 An appropiate approach--neighborhood approach--is proposed forng an approximate agreement in massively prallel systems. Instead-to-all broadcasting that conventional approaches adopt, pproach requires only direct neighbor communication and e routing and redirection. Besides, the approach inherents ties such as isotropic and autonomy, which are very desirabledistributed systems.ealize neighborhood approach, we present a simple but effective m to achieve an approximate agreement on hypercube and wraparoundystems. A set of recurrence relations are derived for each ing to the behavior of neighboring average algorithm. Convergenceor each topology is then deducted through solving the recurrenceons. The evaluated convergence rate, expressed as a function ofure parameters, shows how it is affected by topology of the ch neighboring average algorithm is applied.neighboring average algorithm is then employed to synchronizeand its performance is compared to three representitive globalthms. The performance comparison shows that, besides the advant-n communication and implementation, neighboring average algori-es outperform any global algorithm in some communication Chung-Ping Chung 鍾崇斌 1994 學位論文 ; thesis 86 en_US
collection NDLTD
language en_US
format Others
sources NDLTD
description 博士 === 國立交通大學 === 資訊工程研究所 === 82 === An appropiate approach--neighborhood approach--is proposed forng an approximate agreement in massively prallel systems. Instead-to-all broadcasting that conventional approaches adopt, pproach requires only direct neighbor communication and e routing and redirection. Besides, the approach inherents ties such as isotropic and autonomy, which are very desirabledistributed systems.ealize neighborhood approach, we present a simple but effective m to achieve an approximate agreement on hypercube and wraparoundystems. A set of recurrence relations are derived for each ing to the behavior of neighboring average algorithm. Convergenceor each topology is then deducted through solving the recurrenceons. The evaluated convergence rate, expressed as a function ofure parameters, shows how it is affected by topology of the ch neighboring average algorithm is applied.neighboring average algorithm is then employed to synchronizeand its performance is compared to three representitive globalthms. The performance comparison shows that, besides the advant-n communication and implementation, neighboring average algori-es outperform any global algorithm in some communication
author2 Chung-Ping Chung
author_facet Chung-Ping Chung
Ren-Liang Cheng
鄭仁亮
author Ren-Liang Cheng
鄭仁亮
spellingShingle Ren-Liang Cheng
鄭仁亮
The Approximate Agreement Problem of Massively Parallel Systems
author_sort Ren-Liang Cheng
title The Approximate Agreement Problem of Massively Parallel Systems
title_short The Approximate Agreement Problem of Massively Parallel Systems
title_full The Approximate Agreement Problem of Massively Parallel Systems
title_fullStr The Approximate Agreement Problem of Massively Parallel Systems
title_full_unstemmed The Approximate Agreement Problem of Massively Parallel Systems
title_sort approximate agreement problem of massively parallel systems
publishDate 1994
url http://ndltd.ncl.edu.tw/handle/69764168854156157955
work_keys_str_mv AT renliangcheng theapproximateagreementproblemofmassivelyparallelsystems
AT zhèngrénliàng theapproximateagreementproblemofmassivelyparallelsystems
AT renliangcheng dàliàngpíngxíngxìtǒngzhījìnshìtóngyìwèntí
AT zhèngrénliàng dàliàngpíngxíngxìtǒngzhījìnshìtóngyìwèntí
AT renliangcheng approximateagreementproblemofmassivelyparallelsystems
AT zhèngrénliàng approximateagreementproblemofmassivelyparallelsystems
_version_ 1718351593614606336