具有失效恢复机制的云资源调度算法

发布时间：2018-05-31 15:05

本文选题：云计算 + 贝叶斯估计　；参考：《浙江大学学报(工学版)》2015年12期

【摘要】：针对云服务可靠性较低的问题,提出一种考虑节点失效恢复机制的任务调度模型.该模型引入失效恢复机制分析节点的行为特性,将节点间的交互失效划分为可恢复失效和不可恢复失效.参考社会学的人际关系信任模型,通过量化和评估失效恢复机制下节点的可信程度,建立更加符合实际的云服务可靠性模型,并允许资源节点自行调节失效恢复次数限制和失效恢复率.将节点的可信度并入DLS算法得到考虑失效恢复机制的动态级调度(FR-DLS)算法.FR-DLS算法在计算调度级别时充分考虑服务资源的可信程度,使应用任务能够被有效地分配到可信资源节点上.为了评估所提出的算法,在PlanetLab环境中设计基于CloudSim的仿真实验平台,分析及仿真实验结果表明:所提出的FR-DLS算法在牺牲较少的任务完成时间和调度长度的前提下,能够有效地提高云环境下执行任务的成功率;当云环境中的资源节点数和应用任务数不断增加时,该算法在可靠性方面所提升的性能远高于其在任务完成时间和调度长度代价方面所提升的性能,充分体现了其在大规模云环境下的实用性.
[Abstract]:In order to solve the problem of low reliability of cloud service, a task scheduling model considering node failure recovery mechanism is proposed. In this model, failure recovery mechanism is introduced to analyze the behavior of nodes, and the interactive failure between nodes is divided into recoverable failure and non-recoverable failure. Referring to the interpersonal relationship trust model of sociology, by quantifying and evaluating the trust degree of the node under the failure recovery mechanism, a more realistic cloud service reliability model is established. It also allows resource nodes to adjust the limit of the number of failure recovery and the rate of failure recovery. The credibility of nodes is incorporated into the DLS algorithm to get the dynamic level scheduling algorithm, FR-DLSs, which takes into account the failure recovery mechanism. The FR-DLS algorithm takes full account of the trustworthiness of the service resources when calculating the scheduling level. Enable application tasks to be effectively assigned to trusted resource nodes. In order to evaluate the proposed algorithm, a simulation experiment platform based on CloudSim is designed in PlanetLab environment. The analysis and simulation results show that the proposed FR-DLS algorithm is based on less task completion time and scheduling length. Can effectively improve the success rate of executing tasks in the cloud environment; when the number of resource nodes and application tasks in the cloud environment is increasing, The improved performance of the algorithm in reliability is much higher than that in terms of task completion time and scheduling length cost, which fully reflects its practicability in large-scale cloud environments.
【作者单位】：安徽大学计算机科学与技术学院;铜陵学院数学与计算机科学系;
【基金】：国家自然科学青年基金项目(61402005)
【分类号】：TP393.01

【相似文献】