Research On Online Reliability Prediction Of Component Systems Via Maximum Entropy Markov Model

With the rapid development of information technology,users' demand for software system's business functions is becoming more and more diverse,complicated and varied.Therefore,the traditional development method which is tightly coupled and self-closed is difficult to adapt quickly to the changes in demand of the market or customers.System of Systems(SoS),which is based on service composition,can effectively deel with the challenge.It considers the system as a service and integrates multiple member systems into a new system SoS.Through the idea of service-oriented,loose coupling and cross platform characteristics can be achieved among member systems(component systems).Each component system can be distributed,managed and maintained by itself.The performance of the component system may fluctuate at any time because of the complex and changeable running state and external environment of the component system,which will affect the running of the SoS.Therefore,the quality assurance of SoS is particularly significant.The online reliability prediction technology of SoS component system based on service composition is used to predict the reliability of the component system in the near future.The prediction results can find errors and correct them in time so as to ensure that SoS can run continuously and healthily.However,because the environment of the component system(such as network environment,security environment)is complex and changeable,and the uncertainty of its own running environment(including software and hardware environment),the error of component system appears indistinct regularity.Therefore,this dynamic and uncertain reliability prediction is a challenge to be solved urgently.As for the dynamic and uncertain component system reliability prediction problem,this paper developes the traditional Maximum Entropy Markov Model(MEMM).We introduce the concept of time series motifs,and propose a new prediction model(m_MEMM)which is motifsbased MEMM.Through the historical system parameters of the component system,the prediction model is trained,Based on the model,a complete set of prediction method is presented.Finally,comprehensive experiments are conducted to verify and analyze the correctness and accuracy of the prediction model and method proposed in this paper.