資源描述:
《基于bp算法網(wǎng)格資源調(diào)度的研究》由會(huì)員上傳分享,免費(fèi)在線閱讀,更多相關(guān)內(nèi)容在學(xué)術(shù)論文-天天文庫(kù)。
1、哈爾濱理工大學(xué)碩士學(xué)位論文基于BP算法的網(wǎng)格資源調(diào)度研究姓名:呂昌國(guó)申請(qǐng)學(xué)位級(jí)別:碩士專業(yè):計(jì)算機(jī)應(yīng)用技術(shù)指導(dǎo)教師:孫名松20070301哈爾濱理工大學(xué)工學(xué)碩士學(xué)位論文基于BP算法的網(wǎng)格資源調(diào)度研究摘要網(wǎng)格是新一代的互聯(lián)網(wǎng),資源調(diào)度是網(wǎng)格系統(tǒng)的最核心組成部分,由于網(wǎng)格上的資源具有分布性、異構(gòu)性、動(dòng)態(tài)性等特點(diǎn),使得網(wǎng)格中資源共享的實(shí)現(xiàn)比以前的系統(tǒng)更加困難。而神經(jīng)網(wǎng)絡(luò)是一門模仿人類神經(jīng)中樞——大腦構(gòu)造與功能的智能科學(xué)。具有卓越的自組織、自學(xué)習(xí)能力;善于在復(fù)雜環(huán)境下,快速獲得滿足多種約束條件問題的最優(yōu)化答案,所以把神經(jīng)網(wǎng)絡(luò)引入到網(wǎng)格的資源調(diào)度當(dāng)中,可
2、以很好的發(fā)揮神經(jīng)網(wǎng)絡(luò)的優(yōu)勢(shì),更好的解決網(wǎng)格的資源調(diào)度問題。本文基于網(wǎng)格的基本概念、特點(diǎn),研究了目前比較流行的網(wǎng)格體系結(jié)構(gòu),討論了典型的網(wǎng)格技術(shù)與資源調(diào)度系統(tǒng),-對(duì)兩種主要網(wǎng)格資源模型做了深入分析的基礎(chǔ)上,提出了分層的資源調(diào)度模型和一個(gè)結(jié)構(gòu)簡(jiǎn)單功能完整的資源調(diào)度系統(tǒng)結(jié)構(gòu),把前向反饋神經(jīng)網(wǎng)絡(luò)模型中的BP算法應(yīng)用到網(wǎng)格資源調(diào)度當(dāng)中,詳細(xì)闡述了BP算法在網(wǎng)格資源調(diào)度中的具體應(yīng)用。在比較了幾種常用的網(wǎng)格仿真工具的基礎(chǔ)上,選擇了GridSim這一網(wǎng)格建模與仿真工具箱。在GridSim-I-具箱的仿真環(huán)境下,運(yùn)用JAvA編程語言,對(duì)基于BP算法和優(yōu)先級(jí)算法的
3、資源調(diào)度進(jìn)行了仿真實(shí)驗(yàn)。仿真實(shí)驗(yàn)結(jié)果表明:基于BP算法網(wǎng)格資源調(diào)度的結(jié)果與優(yōu)先級(jí)算法的結(jié)果相比,BP算法的任務(wù)響應(yīng)時(shí)間較快,分配的任務(wù)更加合理,能夠高效的利用網(wǎng)格計(jì)算資源。關(guān)鍵詞網(wǎng)格;資源調(diào)度:BP算法;C_rridSim工具箱蘭塵鎏耋三奎蘭三蘭罌圭主竺蘭蘭ResearchonGridResourceSchedulingBasedOnBPAlgorithmAbstractThegridisanewgenerationofIntemet,gridresourcemanagementandschedulingalethegridsystemmost
4、coreconstituents.Becausegridresourceshasthedistribution,theisomerism,dynamicandSOonthecharacteristics,causedgridtheresourcessharingrealizationtobemoredifficultthanthebeforesystem.111eneuralnetworkisoneimitatesthehumanitynervecenter·-·-cerebrumstructureinthefunctionintelligen
5、tscience.Excellentself-organization,self-learningability;Goodatfastobtainingtheoptimizedanswerwhichsatisfiesmanykindsofconstraintconditionquestionundercomplexenvironment;therefore,theneuralnetworkisintroducedtothegridresourcescheduling.TheneuralnetworkCanplaygoodadvantage,ab
6、ettersolutiongridresourcescheduling.Basedongridbasicconcept,.characteristic,thisthesisstudiedthepresentquitepopulargridarchitecture,discussedtheclassic面dtechnologyandtheschedulingofresourcessystem,analyzedtwokindsofmaingridresourcemodels,proposedthelayeredschedulingofresourc
7、esmodel,proposedonesimpleandfunctionintegrityschedulingofresourcessystemstructure,proposedapplyingthefrontfeedbackneuralnetworkmodeIinBP(BackPropagation)algorithmonthegridschedulingofresources,glaboratedemphaticallybasedonthencuralnetworkBPalgorithmgridresourcemanagementsyst
8、em.IchoseGridSimthisgridsimulationtoolbox,aftercomparingseveralkindsofcommo