云计算环境下资源分配和科学工作流调度的优化方法

发布时间：2021-04-17 18:01

　　科学工作流是多个结构化活动和细粒度计算任务组成的集合,随着科研信息化的出现,工作流调度作为核心组件,已用于描述复杂的多重依赖型任务以及任务之间控制流的表示。与其他应用类似,科学工作流也得益于基础设施即服务云（IaaS云）,基于需求而弹性提供的可扩展性资源可经由IaaS云访问和获取并按需付费,然而大数据应用在云服务资源上的高效调度仍然面临众多亟待解决的挑战,科学工作流任务之间芜杂的关联性增加了问题的复杂度,这促使研究者们不断探寻启发式、元启发式和混合的方法,以期找到工作流调度问题的最优解,因为低效的资源分配和调度除了导致时间和开销增加外毫无裨益。本文针对云计算环境下资源分配和工作流调度提出新的优化方法,这些优化方法使得用户和云服务提供商的许多服务质量（Qo S）指标均得到优化,比如运行时长、运行开销、负载均衡以及资源利用率。本文的主要工作包括:（i）定义和描述云计算环境下调度算法必须解决的问题;（ii）对云计算环境下科学工作流调度问题目前最前沿的先进技术和方法进行全面的分类和总结;（iii）针对在动态可扩展虚拟机组上的任务调度问题,结合启发式和混合整数规划（MIP）模型,提出一种满足最新... 

【文章来源】：华南理工大学广东省 211工程院校 985工程院校 教育部直属院校

【文章页数】：122 页

【学位级别】：博士

【文章目录】：
摘要
ABSTRACT
CHAPTER 1 INTRODUCTION
    1.1 Introduction
    1.2 Background
        1.2.1 Cloud computing
        1.2.2 Scientific workflows
    1.3 Problem definition
    1.4 Motivation
    1.5 Thesis contributions
    1.6 Thesis overview
CHAPTER 2 WORKFLOW SCHEDULING IN THE CLOUD
    2.1 Introduction
    2.2 Scientific workflow scheduling problem in cloud
    2.3 Taxonomies of cloud workflow scheduling problem and techniques
        2.3.1 Computing Environment
        2.3.2 Application model
        2.3.3 Scheduling technique
        2.3.4 Scheduling objective
        2.3.5 Optimization criteria
        2.3.6 Optimization method
        2.3.7 Workload type
        2.3.8 Resource provisioning
        2.3.9 Pricing model
    2.4 Analysis of the workflow scheduling schemes in the cloud
    2.5 Conclusion
CHAPTER 3 SCIENTIFIC WORKFLOW SCHEDULING UNDER DEADLINE CONSTRAINTS
    3.1 Introduction
    3.2 Related work
    3.3 System model
        3.3.1 Scientific workflow application model
        3.3.2 Cloud resource model
        3.3.3 Workflow execution model
    3.4 The proposed DSB workflow scheduling algorithm
        3.4.1 Assumptions
        3.4.2 Problem statement
        3.4.3 Basic definitions
        3.4.4 Proposed algorithm
        3.4.5 Computational complexity
    3.5 Performance analysis and discussion
        3.5.1 Experiment environment
        3.5.2 Performance metric
        3.5.3 Evaluation results
        3.5.4 Sensitivity of overheads, VM performance variations and task failures
        3.5.5 Analysis of Variance （ANOVA） test
    3.6 Conclusions
CHAPTER 4 HYBRID METAHEURISTIC FOR MULTI-OBJECTIVE WORKFLOW SCHEDULING
    4.1 Introduction
    4.2 Related work
    4.3 Problem description for the proposed methodology
        4.3.1 System model
        4.3.2 Assumptions
        4.3.3 Multi-objective optimization
        4.3.4 Problem formulation
    4.4 Proposed work
        4.4.1 Initialization
        4.4.2 Fitness evaluation
        4.4.3 Optimization
        4.4.4 Selection of best fit solutions
        4.4.5 Termination condition
    4.5 Performance evaluation
        4.5.1 Experimental setup
        4.5.2 Evaluation metrics
        4.5.3 Simulation results
        4.5.4 Analysis of Variance （ANOVA） test
    4.6 Conclusions and future work
CHAPTER 5 Structure-aware and budget-aware workflow scheduling
    5.1 Introduction
    5.2 Related work
    5.3 Models and problem definition
        5.3.1 Application model
        5.3.2 Cloud model
        5.3.3 Definitions
        5.3.4 Problem formulation
    5.4 Proposed algorithm
        5.4.1 Task ranking
        5.4.2 DAG Partitioning
        5.4.3 Task grouping
        5.4.4 Budget propagation
        5.4.5 Dynamic resource provisioning and scheduling
    5.5 Performance evaluation
        5.5.1 Experimental setup
        5.5.2 Performance metric
    5.6 Evaluation results
    5.7 Conclusion
CONCLUSIONS AND FUTURE WORK
    Conclusions
    Future work
REFERENCES
攻读博士学位期间取得的研究成果
ACKNOWLEDGEMENTS
附件



本文编号：3143891