| 本文已被：浏览 3551次 下载 4966次 |
码上扫一扫！ |
|
|
| 一种云环境中数据流的高效多目标调度方法 |
|
沈尧1, 秦小麟1,2, 鲍芝峰3
|
|
1.南京航空航天大学计算机科学与技术学院, 江苏 南京 210016;2.软件新技术与产业化协同创新中心, 江苏 南京 210016;3.School of Computer Science & Information Technology, RMIT University, Melbourne VIC 3000, Australia
|
|
| 摘要: |
| 在分布式系统中，云计算作为一种新的服务提供模式出现，其执行科学应用数据流时的优势和缺点得到越来越多的关注，其主要特点为拥有大量同质和并发的任务包，并构成了性能瓶颈的主要因素.在云数据流中调度大规模任务是已被证实的NP难问题.专注于解决优化云数据流中的调度过程，并受现实世界启发，从不同角度将优化目标分别划分为用户指标（完工时间和经济成本）和云系统指标（网络带宽、存储约束和系统公平度），并将该调度问题制定成为一个新的连续的合作博弈，设计出快速收敛的高效Muliti-Objective Game（MOG）调度算法，在优化用户指标的同时，实现系统指标的约束，并保证云资源的效率和公平度.通过综合实验，证实该方法与其他相关算法相比，在算法复杂度O（l·K·M）（明显改进数量级）、结果质量（一些情况下最佳）、系统级别公平性上具有明显的优越性. |
| 关键词: 云计算 多目标优化 合作博弈 数据流处理 任务调度 |
| DOI：10.13328/j.cnki.jos.005158 |
| 分类号:TP311 |
| 基金项目:国家自然科学基金（61373015，61300052）；中央高校基本科研业务费专项基金（NP2013307）；江苏省研究生培养创新工程（KYLX_0287） |
|
| Effective Multi-Objective Scheduling Strategy of Dataflow in Cloud |
|
SHEN Yao1, QIN Xiao-Lin1,2, BAO Zhi-Feng3
|
|
1.College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics, Nanjing 210016, China;2.Collaborative Innovation Center of Novel Software Technology and Industrialization, Nanjing 210016, China;3.School of Computer Science & Information Technology, RMIT University, Melbourne VIC 3000, Australia
|
| Abstract: |
| As a new emerging service provider, cloud computing, exhibiting advantages and disadvantages when executing the scientific data flows, is getting more and more attention. One of the main factors that constitute the performance bottleneck is there are many homogeneous and concurrent task packages in cloud. This paper focuses on optimizing the scheduling process in dataflow and transforming the optimization objectives into user metrics (makespan and economic cost) and indicators of cloud systems (network bandwidth, storage constraints and system fairness). An efficient multi-objective game algorithm (MOG) is proposed by formulating the optimization problem as a new cooperative game. The MOG method is able to optimize the user metrics while satisfying the constraint of the system metrics and ensuring the efficiency and fairness of the cloud resources. Comprehensive experiments demonstrate that compared with other related algorithms, the proposed MOG method has obvious advantages in terms of algorithm complexity O(l·K·M) (improvement of magnitude), result quality (optimum in some cases) and system level fairness. |
| Key words: cloud computing multi objective optimization cooperative game dataflow processing task scheduling |
