【摘要】：大規(guī)模高速緩存一致性非均勻存儲(chǔ)訪問(cache coherence non-uniform memory access,CC-NUMA)系統(tǒng)通常采用兩級一致性域方法來降低緩存一致性協(xié)議維護(hù)開銷,提升系統(tǒng)性能.兩級一致性域系統(tǒng)中,多個(gè)處理器互連,形成結(jié)點(diǎn)內(nèi)一致性域;多個(gè)結(jié)點(diǎn)互連,形成結(jié)點(diǎn)間一致性域.然而,受限于處理器直連能力與處理器可識(shí)別ID數(shù),系統(tǒng)的單結(jié)點(diǎn)規(guī)模有限,系統(tǒng)規(guī)模的擴(kuò)展不得不依靠增加結(jié)點(diǎn)數(shù)來實(shí)現(xiàn),使得大規(guī)模CC-NUMA系統(tǒng)的結(jié)點(diǎn)間互連復(fù)雜度上升,跨結(jié)點(diǎn)訪問帶寬和延遲急劇增長,影響了系統(tǒng)性能的有效擴(kuò)展.MPD系統(tǒng)通過在結(jié)點(diǎn)內(nèi)構(gòu)建多個(gè)并行緩存一致性域,突破了處理器直連能力與可識(shí)別ID數(shù)對單結(jié)點(diǎn)規(guī)模的限制,能夠大幅減少結(jié)點(diǎn)數(shù)量,并將部分結(jié)點(diǎn)間訪問轉(zhuǎn)化為結(jié)點(diǎn)內(nèi)訪問,實(shí)現(xiàn)系統(tǒng)性能的有效擴(kuò)展.理論分析和實(shí)驗(yàn)結(jié)果表明:采用同規(guī)格處理器的32路系統(tǒng)中,結(jié)點(diǎn)內(nèi)4個(gè)并行緩存一致性域的MPD系統(tǒng)可實(shí)現(xiàn)結(jié)點(diǎn)數(shù)目減少75%、一致性目錄存儲(chǔ)開銷節(jié)省40%以上、平均訪問延遲降低約27.9%、系統(tǒng)整體性能提升約14.4%.
[Abstract]:Large-scale cache consistent non-uniform storage access (cache coherence non-uniform memory access,CC-NUMA) systems usually use a two-level consistency domain approach to reduce cache consistency protocol maintenance overhead and improve system performance. In a two-level uniform domain system, multiple processors interconnect to form a consistent domain within a node, and multiple nodes interconnect to form a consistent domain between nodes. However, limited by the direct connection ability of the processor and the number of ID identifiable by the processor, the single node size of the system is limited, and the expansion of the system scale has to be realized by increasing the number of nodes, which increases the complexity of interconnection between nodes in large-scale CC-NUMA systems. The rapid increase of cross-node access bandwidth and delay affects the effective expansion of system performance. MPD system breaks through the limitation of processor direct connection ability and identifiable ID number to single node size by building multiple parallel cache consistency domains in nodes. The number of nodes can be greatly reduced, and the access between some nodes can be transformed into intra-node access, so as to realize the effective expansion of system performance. The theoretical analysis and experimental results show that in the 32-channel system with the same specification processor, the number of nodes can be reduced by 75% and the storage cost of the consistent directory can be saved by more than 40% in the MPD system with four parallel cache consistent domains in the node. The average access delay is reduced by about 27.9%, and the overall performance of the system is improved by about 14.4%.
【作者單位】： 高效能服務(wù)器和存儲(chǔ)技術(shù)國家重點(diǎn)實(shí)驗(yàn)室(浪潮集團(tuán)有限公司);
【分類號】：TP303

【相似文獻(xiàn)】

相關(guān)會(huì)議論文 前1條

1 董小潔;夏寬理;;忍受犯規(guī)的開放式提交協(xié)議[A];第十二屆全國數(shù)據(jù)庫學(xué)術(shù)會(huì)議論文集[C];1994年

相關(guān)重要報(bào)紙文章 前1條

1 蔣杰 方力 竇文華;覆蓋控制[N];計(jì)算機(jī)世界;2004年

相關(guān)博士學(xué)位論文 前1條

1 王立;城市結(jié)點(diǎn)文化特質(zhì)及其協(xié)同觀[D];重慶大學(xué);2006年

相關(guān)碩士學(xué)位論文 前6條

1 楊孟君;基于網(wǎng)絡(luò)認(rèn)知的無中心式系統(tǒng)交互的優(yōu)化方法[D];電子科技大學(xué);2015年

2 田艷;復(fù)雜網(wǎng)絡(luò)重要結(jié)點(diǎn)發(fā)現(xiàn)算法研究[D];云南財(cái)經(jīng)大學(xué);2016年

3 姜浩;對等網(wǎng)絡(luò)中路由中繼結(jié)點(diǎn)發(fā)現(xiàn)機(jī)制的研究[D];華中科技大學(xué);2007年

4 顧燁;P2P網(wǎng)絡(luò)邏輯拓?fù)鋬?yōu)化和結(jié)點(diǎn)組管理策略研究[D];浙江工商大學(xué);2010年

5 李丁丁;一種基于結(jié)點(diǎn)聚類的網(wǎng)絡(luò)定位算法[D];湖南大學(xué);2008年

6 孫宇奇;基于復(fù)雜網(wǎng)絡(luò)的社團(tuán)發(fā)現(xiàn)研究[D];遼寧師范大學(xué);2011年

，

本文編號：2485545