服务系统在线可靠性预测方法及关键技术研究

Service-orineted System of Systems (SoS) considers System as a Service. It aims to construct a robust and value-added SoS based on service composition; it can therefore be regarded as a type of Service-Oriented System. To cope with the dynamic and uncertainty running environment and assure the software qualiy, the online reliability time series prediction for SoS has been a grand challenge for the technologies of Service-Orineted Computing. This project aims to investigate the methodologies for predicting component level and system level SoS’s online reliability time series. It aims to provide a support for proactive fault removal that SoS can adapt to complex and dynamic environments; henceforce provide a mechanism to guarantee the stable execution of SoS. Our research will be conducted from multiple perspectives and is expected to deliver the following results: (1) methods for component level online reliability time series prediction via Probabilistic Graphical Models; (2) Fluctuation RAte interval Mapping function based Interval Temporal Logic (FRAM-ITL) model and its application for component level online reliability time series prediction; (3) methods for calculating the system level reliability under dependency and independency system invocations; (4) the evaluation of these methods and fault removal mechanisms will be conducted in a real Air Trafic Control (ATC) system. We believe our research will be a significant contribution to the research and application of software quality assurance for Service-Oriented Systems and the advance of SoS technologies.

面向服务的“系统的系统”（System of Systems, SoS）采用服务化思想，将系统视为服务，通过服务组合构造一个新的系统。所构建的服务系统（Service-Oriented Systems）运行在动态、不确定的环境下。为保障系统持续有效运行，实时预测系统未来的可靠性以提前移除系统错误，成为SoS研究和应用中需要解决的挑战性问题。本项目围绕服务系统可靠性在线时间序列预测问题开展研究，分别预测SoS组件级（组件系统）及系统级可靠性，提出：（1）基于概率图模型的组件系统可靠性在线时间序列预测方法；（2）基于波动率映射函数的区间时态逻辑模型及组件系统可靠性在线时间序列预测方法；（3）能够同时支持依赖型及非依赖型调用模式的SoS系统级可靠性计算方法。在上述预测的基础上，面向空管SoS软件质量保障问题，开展应用研究。项目研究对开展基于服务组合的大规模复杂软件系统的研究和应用具有重要的意义。

面向服务的“系统的系统”（System of Systems, SoS）采用服务化思想，将系统视为服务，通过服务组合构造一个新的系统。所构建的服务系统（Service-Oriented Systems）运行在动态、不确定的环境下。为保障系统持续有效运行，实时预测系统未来的可靠性以提前移除系统错误，成为SoS研究和应用中需要解决的挑战性问题。本项目围绕服务系统可靠性在线时间序列预测问题开展研究，分别预测SoS组件级（组件系统）及系统级可靠性。本项目主要研究了三个方面的内容：1）采用基于时间序列的动态贝叶斯网络和多步轨迹的动态贝叶斯网络实现了面对不确定性错误事件的SoS组件系统可靠性时间序列在线预测；2）采用基于motifs的最大熵马尔可夫模型实现了随机波动环境下SoS组件系统可靠性在线预测；3）采用一种多簇自适应脑风暴优化算法实现了依赖型/非依赖型调用模式下SoS系统级可靠性的在线预测。最后将以上技术在中电28所的空管系统上进行了验证。本项目研究对开展基于服务组合的大规模复杂软件系统的研究和应用具有重要的意义。