基于审查不确定性的预见式软件自适应

Proactive self-adaptation of software based on inspecting uncertainty

为了解决传统的被动式自适应(RSA)模型自适应不可预期的问题,提出预见式的自适应(PSA)方法.PSA方法要解决的重要问题是:模型如何能够从环境中学习.通过使用隐Markov模型(HMM),系统能够从历史行为中进行学习并生成预见式的动作.和传统的RSA模型相比,PSA方法能预见式地调整系统的运行时行为以适应新的环境.通过对系统管理的观察和PSA方法的认知能力,应用系统能够做出合理的决策.应用程序能实现预见性的自主管理过程,减少了人工干预.实验结果表明,PSA方法为应用提供了预见式的自适应管理机制,提高了应用的可管理性和服务质量(QoS).

A method of proactive self-adaptation （PSA） was proposed to address the unanticipated adaptation of the traditional reactive self-adaptation （RSA） model.The PSA method presented an important problem to be resolved how the model learns from the environment.Hidden Markov model （HMM） was employed to learn from history behavior of target system,and then generated anticipatory actions.The PSA method can proactively adjust the runtime behaviors of the system to be adaptive to the new situations compared to the traditional RSA model.The application system made sound decision by combining the observation from system administrators and the cognitive power of PSA.Then applications implemented the proactive autonomic management and reduced manual operation.Experimental results show that the PSA method provides for application with proactive self-adaptive management mechanism and improves the manageability and quality of service （QoS） of application.