The open and distributed service platform is prone to containing more management functions and supporting stronger decentralized interaction, so the hard-management and high-cost problems become more serious. For that, a self-managing service platform architectural reference model, characterized by component-based, service-oriented and interaction-centric, is introduced in this paper. On this ground, a hierarchical autonomic control structure based on feedback mechanism is proposed, which takes the service components and their interaction relations as controlled objects, and makes decision according to the analytical performability model of service platform. A combination of availability and performance models of management service is suggested via Markov Procedure, Stochastic Petri Net and Queueing Network theory. Moreover, the ability of failure recovery into the service components and links are taken into account. Numerical examples show that the proposed performability model based on Queueing Petri Net can quantitatively analyze the influence of failure probability and recovery time on availability and performance of service platform, and autonomic control approach is helpful for improving the effectiveness of service platform.%开放的分布式服务平台倾向于涵盖更丰富的管理功能,支持更强的分散交互性,从而导致软件管理和维护的难度和成本问题日益突出.为此,引入一种自管理的服务平台体系结构参考模型,以构件作为功能实现载体,服务作为功能组织手段,交互作为功能扩展方式.提出了一种基于分层反馈的自主控制架构,以服务构件及相互之间的交互关系为控制对象,执行力模型为决策基础.在管理服务的可用性和性能建模中运用了马尔可夫过程、随机Petri网和排队网模型理论,并考虑了链路和节点的失效修复机制.仿真结果表明,基于排队Petri网的执行力模型能够反映失效率和修复时间对服务平台性能和可用性的影响,并验证了自主控制方法对提高服务平台有效性的积极作用.
展开▼
