高级检索
    周墨颂, 董小社, 陈衡, 张兴军. 基于计算资源运行时剩余能力评估优化云平台[J]. 计算机研究与发展, 2017, 54(11): 2516-2533. DOI: 10.7544/issn1000-1239.2017.20160700
    引用本文: 周墨颂, 董小社, 陈衡, 张兴军. 基于计算资源运行时剩余能力评估优化云平台[J]. 计算机研究与发展, 2017, 54(11): 2516-2533. DOI: 10.7544/issn1000-1239.2017.20160700
    Zhou Mosong, Dong Xiaoshe, Chen Heng, Zhang Xingjun. Improving Cloud Platform Based on the Runtime Resource Capacity Evaluation[J]. Journal of Computer Research and Development, 2017, 54(11): 2516-2533. DOI: 10.7544/issn1000-1239.2017.20160700
    Citation: Zhou Mosong, Dong Xiaoshe, Chen Heng, Zhang Xingjun. Improving Cloud Platform Based on the Runtime Resource Capacity Evaluation[J]. Journal of Computer Research and Development, 2017, 54(11): 2516-2533. DOI: 10.7544/issn1000-1239.2017.20160700

    基于计算资源运行时剩余能力评估优化云平台

    Improving Cloud Platform Based on the Runtime Resource Capacity Evaluation

    • 摘要: 云平台资源管理中存在资源供给与需求不匹配的问题,导致平台性能受到严重影响.针对此问题,基于相似任务建立运行时计算资源剩余能力评估模型,该模型利用云计算负载中相似任务执行逻辑相同的特点,使用相似任务代替测试程序量化资源剩余能力,避免了执行测试程序的计算资源代价;依据该模型提出了一种运行时云计算资源剩余能力分类评估方法RCE(resource capacity evaluation),该方法综合各方面因素评估运行时资源剩余能力,具有运行时代价低、评估结果准确且有时效性的特点.将RCE评估结果应用在若干算法中,以提高云平台资源供给与需求的匹配程度并优化云平台各方面性能;在独享环境和真实云环境中验证了RCE方法和基于RCE的算法,实验结果表明:RCE评估结果及时反映了计算资源能力变化,为算法和平台的优化提供了有力支持,基于RCE优化的算法解决了云计算资源管理中资源供给与需求不匹配问题并大幅提高云计算平台性能.

       

      Abstract: There is a mismatch between computing resource supply and demand in cloud computing platform resource management, which leads to the performance degradation. This paper establishes a runtime computing resource available capacity evaluation model base on similar tasks. The model uses the characteristic of cloud computing workload in which similar tasks have the same execution logic, evaluates computing resource available capacity according to similar tasks avoiding computing resource consumption in executing benchmark. This paper applies the model to propose a computing resource capacity evaluation method called RCE, which considers many factors and evaluates runtime computing resource available capacity classified by resource type. This method gets accurate evaluation results timely with little cost. We apply RCE results in some algorithms to match computing resource supply and demand, and improve cloud computing platform performance. We test RCE method and algorithms base on RCE in dedicated and real cloud computing environments. The test results show that the RCE method gets runtime evaluation results timely and the evaluation results reflect computing resource available capacity accurately. Moreover, the RCE method supports the optimization of algorithm and platform effectively. And algorithms base on RCE resolve the mismatch problem between resource supply and demand, and significantly improve the performance of cloud computing platform.

       

    /

    返回文章
    返回