Research And Application Of QoS-based Adaptive Control Middleware

With the maximization and complication of software system, the desire of software system performance is higher. The development of Automatic Control is to enlarge the applications, for example, control theory and method is applied to the software system design. In the complicated applications, to provide the effective method and tools to support QoS guarantee and user-oriented Differentiated Service is a challenging and critical problem. The main work of this thesis is as follows:First, a QoS-based adaptive control middleware architecture is designed. It can provide performance guarantee for software system. The software development process based on adaptive control middleware is introduced. We describe the detailed process including mapping software performance guarantee and classified QoS to adaptive control problem, composition of control loop components, system identification, controller parameter tuning and implement of adaptive control. The adaptive control middleware with 'online identification, online tuning' feature can improve the modeling accuracy and control effect of software system.The RLS-AFF (Recursive Least Square-Adaptive Forgetting Factor) online identification algorithm is presented. The model structure and parameters are identified by RLS offline identification. RLS-AFF uses identified parameters as initial parameters of online identification and corrects them further. The experiment result shows that identification accuracy and convergence rate of RLS-AFF is promoted. The parameters of Web Services Composition Execution Engine identified by RLS-AFF can reflect the variety track of the engine in dynamic running process, thus the dynamic parameter track is achieved. Accelerated Simplex Method is adopted in controller parameters self-tuning. It includes some features, such as rapid convergence rate of controller parameter, simplicity and utility, and capability of online optimization.The adaptive control middleware is used in Web Services Composition Execution Engine. The thesis gives the definition of service response time guarantees in Web Services Composition Execution Engine. The architecture of execution engine supporting response time guarantees is design and implemented. At run-time the engine implements QoS-based adaptive control and optimization technology. The performance evaluation experiment demonstrates Web Services Composition Execution Engine based on adaptive control middleware can provide good response time guarantees for classified service request even when workload varies significantly.