【摘要】 云计算通过网络共享成功实现了计算资源的高效利用。云资源分配的动态性、随机性、开放性使得云平台的服务质量保障难题日益突出。云环境下资源状态的监控技术通过深入挖掘分析监控数据,及时发现计算资源的异常运行状态。根据历史运行数据等对资源的未来使用状态做出预测。以便及时发现潜在的性能瓶颈和安全威胁,为用户提供可靠稳定的云服务。结合实例介绍了在资源状态监控方面有代表性的研究方法,包括概率分析、方程拟合、聚类分析等方法,并对比了各类方法的性能特点及局限性。最后讨论了云资源状态监控技术在数据复杂度和规模等方面所面临的技术挑战,并指出了原始数据去冗降维、算法设计突出非监督化、计算任务向终端推送、分析结果协同增效等未来发展趋势。更多还原

【Abstract】 Cloud computing successfully achieves the efficient use of computational resources through internet sharing. The characteristics of cloud resources allocation such as the dynamics property,randomness and openness make the difficulty in Qo S( Quality of Service)assurance be increasingly noticeable. Through mining and analysing in depth the monitoring data,the monitoring technologies for resource state in cloud environment find timely the abnormal operation states in those computational resources,and make the prediction on resources usage state in the future according to historical operation data so as to timely discover potential performance bottlenecks and security threats,these provide the reliable and stable cloud services to users. In combination with instances,in the paper we introduce some representative research approaches in regard to resources states monitoring,including probability analysis,equation fitting and clustering analysis,etc.,and compare the performance features and limitations of different methods. In end of the paper,we discuss the technical challenges encountered by the monitoring technologies for cloud resource states in the aspects of data complexity and scale,and point out the future development trend such as redundancy removal and dimensionality reduction of primitive data,non-supervision highlighting in algorithm design and analysis,pushing the computational tasks to terminals,and synergies of analysis results,etc.更多还原