在云上執(zhí)行復(fù)雜的科學(xué)工作流應(yīng)用程序通常會(huì)涉及大量的虛擬機(jī)（VM）,這使得成本和能耗成為人們關(guān)注的焦點(diǎn)。為了緩解該問題,一些云服務(wù)提供商（例如CloudSigma和ElasticHosts）引入了新的定價(jià)策略,根據(jù)分配的CPU頻率以及虛擬機(jī)配置和價(jià)格的各種組合對用戶收費(fèi)。然而,可定制的CPU頻率使資源分配和調(diào)度變得更加困難,難以實(shí)現(xiàn)成本最優(yōu)的調(diào)度方案。較高的CPU頻率會(huì)帶來高能耗并增強(qiáng)可靠性,而為了降低能耗,降低CPU頻率會(huì)產(chǎn)生軟錯(cuò)誤問題,從而導(dǎo)致工作流應(yīng)用程序在規(guī)定時(shí)間內(nèi)完成的失敗率很高。因此,非常需要一種頻率調(diào)節(jié)方法來獲得成本優(yōu)化的工作流調(diào)度解決方案。為了解決上述挑戰(zhàn),本研究提出了一種基于遺傳算法的新方法。該方法采用新引入的遺傳算子（即交叉和變異）,在能耗、可靠性、最大完工時(shí)間和內(nèi)存約束下,將任務(wù)分配到具有特定工作頻率的虛擬機(jī)中,為云工作流快速找到成本最優(yōu)的資源供應(yīng)和任務(wù)調(diào)度解決方案。本論文主要有以下三點(diǎn)貢獻(xiàn):1.在考慮帶回滾恢復(fù)的檢查點(diǎn)開銷的情況下,將最大完工時(shí)間、能耗、內(nèi)存和可靠性約束下的云工作流任務(wù)調(diào)度的成本優(yōu)化問題形式化。2.基于遺傳算法,我們引入了一種新的遺傳算子,即染色體... 

【文章來源】：華東師范大學(xué)上海市 211工程院校 985工程院校 教育部直屬院校

【文章頁數(shù)】：87 頁

【學(xué)位級別】：碩士

【文章目錄】：
摘要
ABSTRACT
Notations used in this thesis
1 Introduction
    1.1 Research Background and Significance
    1.2 Research Challenges and Contributions
    1.3 Relevant Work
        1.3.1 Cost Optimization in Cloud Computing
        1.3.2 Energy Saving Technologies of Data Center
        1.3.3 Cloud Service Fault Tolerance Technology
    1.4 Thesis Organization
2 Introduction to Relevant Theories
    2.1 Cloud Computing
        2.1.1 Classification Based on Service Type
        2.1.2 Classification Based on Deployment Mode
    2.2 Soft Error
    2.3 Checkpointing with Rollback Recovery
    2.4 Summary
3 System Models and Problem Description
    3.1 System Models
        3.1.1 Virtual Machine Model
        3.1.2 Workflow Model
        3.1.3 Fault Tolerant Task Model
        3.1.4 Energy Consumption Model
    3.2 Problem Description
    3.3 Summary
4 Cost Optimizing Workflow Scheduling Algorithm
    4.1 Algorithm Overview
    4.2 Detailed Design of Algorithm
        4.2.1 Chromosome Encoding
        4.2.2 Frequency Selection
        4.2.3 Chromosome Regularization
        4.2.4 Fitness Function
        4.2.5 Crossover
        4.2.6 Mutation
        4.2.7 Chromosome Modification
    4.3 Case Study
    4.4 Summary
5 Experimental Results and Analysis
    5.1 Experimental Setup
        5.1.1 Experimental Parameters
        5.1.2 Workflow Data Set
        5.1.3 Comparison Algorithms
    5.2 Results and Analysis
        5.2.1 Cost Optimization with Different Makespan Goals
        5.2.2 Energy Optimization with Different Makespan Goals
        5.2.3 Cost Optimization with Different Reliability Goals
        5.2.4 Energy Optimization with Different Reliability Goals
    5.3 Summary
6 Conclusion and Future Work
    6.1 Conclusion
    6.2 Future Work
Bibliography
Acknowledgements
Research Outcomes



本文編號(hào)：3572027