| As a new software application mode, Internetware is constructed and operated on the open and dynamic network. It will be an important development and application direction in the future, and its reliability attracts more and more attention. This dissertation mainly studies Internetware reliability mode theory and computation method. The related key technologies are Internetware calculus based on process algebra, reliability analysis and reliability distribution computation based on Markov chain. This dissertation includes six parts.First, the main technical problems and solutions in terms of Internetware reliability are studied and summarized systematically. The background information, the application characteristics and the concept of Internetware reliability and its research status are analyzed and discussed. The current software and Internetware are analyzed comparatively, the key technical problems and solutions in terms of Internetware reliability are studied systematically. The study shows that though Internetware has made great development in the system architecture and the application as a new software mode, there are still many technical problems about reliability which hinder its wide application, such as reliability model construction, reliability computation,transition probability computation, reliability evolution, reliability distribution computation, and so on. This study provides theoretical basis and reference for further research.Second, an Internetware reliability formal method based on process algebra is proposed. The model based on running characteristic of Internetware is established, the representation method based on process algebra is proposed, and thus the evolutionary computation of Internetware reliability is implemented. The quantified expression and computation of Internetware is realized from the perspective of algebra. The formal definition of Internetware and Internetware reliability are given first. It embodies the dynamic characteristics and adaptation of Internetware, and it is used as the basis for analysis. Then the reliability combinatorial operation of Internetware component is proposed, and its algebra system is constructed. Subsequently reliability combinatorial evolutionary computation method is proposed, which solves the problems of the dynamic, adaptive evolutionary computation. The experiment shows that the proposedalgorithm is in accordance with the law of Internetware reliability.Third, a computation method of Internetware reliability is proposed, including component reliability and system reliability. Based on the above research, according to the operation state and the operation characteristics of Internetware, and based on Markov chain and Bayesian theory, the model and computation method of Internetware reliability are presented, and the analysis and computation of Internetware reliability is achieved. Markov chain is constructed on the basis of architecture, and the reliability computation model is established; then, transition probability computation method is put forward, which is the basis of reliability computation; considering network visiting characteristics and its performance, according to Poisson distribution, the computation method of reliability of the connection of Internetware components is proposed, which achieves connection reliability computation in network environment; last, based on probability theory and stochastic theory, transition occupation rate is used to estimate transition probability, and based on gradient method, the computation method of transition probability matrix is proposed. The experiment shows that the proposed algorithm can compute Internetware reliability effectively.Fourth, a computation method of Internetware reliability evolution is proposed.Internetware reliability evolution is an important evolution property. According to convolution theory and evolution properties, the definition of Internetware reliability evolution is given, reliability evolution method and computation method are proposed,which ensure effectiveness from complexity. Meanwhile the reliability evolution computation problem of component reliability simultaneous change in the traditional modes is solved by introducing the reliability vector. The experiment shows that the proposed reliability evolution model and computation method can make reliability evolution analysis and performance assessment effectively.Fifth, based on cascade punishment, a mixed dynamic optimization algorithm of Internetware reliability distribution is proposed. Reliability distribution computation method is proposed based on DTMC(Discrete Time Markov Chain). Reliability computation function formation is analyzed and studied, and reliability function formation method is put forward, which solves the main difficulty that reliability computation is hindered because of path explosion. According to the characteristics of Internetware, reliability cost function is designed, which provides theoretical basis for reliability distribution. Last, based on cascade punishment, GA-PSO mixed dynamicoptimization algorithm is designed, which achieves effective reliability distribution. The experiment shows that the proposed method of reliability distribution can meet the demand of the reliability distribution and the software cost control.Based on the current study, the prospects of Internetware reliability development is also put forward in this dissertation. First, Internetware architecture and new method of transition probability automatic extraction will be researched to make immediate reliability computation and support Internetware reliability analysis effectively. Then,considering network influence factors, more effective reliability computation method will be designed, which can solve much faster and provide more reliability assessment indexes. Finally, reliability computation customization design and parallel computation can help to make different optimization according to different application scene.Combining with the new network technologies and using the rich network computing resources, they help to improve computation effectiveness, and be applied to more complex software effectively.
|
PDF Full Text Request