【摘要】 云环境下资源状态变化趋势预测技术通过深入挖掘分析监控数据,根据历史运行状态以及其他经验规律,对资源的未来使用状态的可能发展趋势作出预测,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变化趋势预测方面有代表性的研究方法,包括概率分析、方程拟合、机器学习、事件感知等,并对比了各类方法的性能特点及局限性;随后,给出云环境资源状态预测研究常用的实验平台、标准数据集合以及评价指标等的简要介绍;最后讨论了云资源状态变化趋势预测技术在数据复杂度和规模等方面所面临的技术挑战。在将来,轻量级、可扩展、非监督的资源状态分析算法将获得更多关注,而融合并利用计算资源自身特征的研究方法将能够更准确地预测虚拟资源的状态变化趋势。更多还原

【Abstract】 By analyzing running data of those computing resources,cloud services provider can detect and predict potential anomalies in the resources to get a whole picture of their running status. At present,most cloud service providers have already opened status monitoring tools to their users as a basic service,while any further conclusion drawing from data analysis is still in the preliminary analysis stage. In resources status predicting area,the current mainstream methods included the probabilistic analysis,equation fitting,machine learning and event-aware methods. This paper introduced the premise assumptions of the above types of methods and analyzed their performance with practical applications. It also discussed the development trend of cloud resources status monitoring and forecasting techniques. In cloud computing environment,lightweight,scalable and unsupervised analysis algorithms will get more attention. More accurate result can be achieved while considering the inherent characteristics in algorithms designing.更多还原