Researches On Self-adaptable Architecture Of Computing System

Self-adaptable system can dynamically adjust its structure and behaviour to deal with the performance degradation. Lack of efficient techniques in guiding software requirement and design leads to difficulties in developing software products and product maintainence, as is suffering in developing self-adaptable system. This thesis investigates essential problems and techniques in constructing software architecture of self-adaptable system, which contributes to improving its flexibility and reliability of software. The main contributions of this thesis are summarized as follows:Firstly, considering that the existing self-adaptable Architecture lack of cohesion, reuse difficult issues, we present C2-style used in self-adaptable architecture composed by components and connectors, and then we apply it to VOD system. This thesis uses stochastic Petri net to analyze the performance of system, uses UML to describe software architecture from some views. In this way, it can reduce the complexity of modeling architecture.Secondly, the existing framework of the self-adaptable system coupling, interactive difficult issues, two kinds of self-adaptable architecture based on Agent are presented: based on multi Agent and mobile Agent. Agent has high independent, sociality and adaptability. So we can take easy to solve how to make system agile and adaptable by negotiating between Agents. Then we present a new self-adaptable architecture based on mobile Agent, because mobile Agent can take some important information from one place to another, which can avoid the congestion of network, as different components often communicate together. When we begin to analyze these architectures, this paper uses extension of G-Net, layered timed CPN to model and analyze.Thirdly, we propose composing redundant structure and rejuvenation schedule for constructing Fault-Tolerant system, and then use Non-Markovian stochastic Petri nets and timed CPN respectively to analyze.We propose an agent-based architecture for adaptive grid computing, then model and formalize the architecture with timed-Coloured Petri Nets.Finally, we propose a service-oriented self-adaptable Instant Message prototype system, in the face of continuously increasing client requests, through the dynamic reconfiguration of system (part of the requests will be forwarded to spare server), reduce system load, then we propose multi-genetic algorithms that control and manage dynamic reconfiguration of system for satisfying some conflicting goals.