Abstract: It focuses on the problem of resources configuration optimization for cloud applications running in cloud environment, proposes a service level objective aware cloud application resources auto-adaptive configuration technique. It presents methods to generate Bayesian network based probabilistic reasoning model by the collected historical metrics data, and uses predefined application quality of service level objectives to find virtual devices in the resource over-provisioning or under-provisioning state, hence generates optimization strategy accordingly. Emulation test environment results show that the request-response time of cloud application with the adaptive reconfiguration optimization algorithm open is better than others.

Key words: cloud application, probabilistic inference, Bayesian network, Service Level Objectives（SLO） assurance

摘要： 针对部署在云环境下的云应用资源配置优化问题，提出了通过感知服务质量变化自适应配置云应用资源的策略。设计了一种基于采集到的历史运行指标数据生成贝叶斯网络概率推理模型，并利用预定义应用服务质量目标（Service Level Objectives，SLO）找到资源超配或不足的虚拟设备，生成资源优化方案。在仿真环境下验证测试结果显示，开启自适应配置算法的云应用请求响应时间指标明显优于未开启的云应用。

关键词: 云应用, 概率推理, 贝叶斯网络, 服务质量目标保障