Font Size: a A A

Online Reliability Prediction Of Service-Oriented Systems Based On Deep Learning

Posted on:2020-04-08Degree:MasterType:Thesis
Country:ChinaCandidate:X LinFull Text:PDF
GTID:2428330623459859Subject:Computer Science and Technology
Abstract/Summary:PDF Full Text Request
With the emergence of the requirements of large-scale complex systems and the rapid development of service computing technology,Service-Oriented Systems based on SOA have emerged.Service-Oriented Systems refer to the combination of services with different functions in a certain way to meet the complex needs of users.Therefore,Service-Oriented Systems need a high degree of collaboration of multiple component services,which requires that each component service must have good performance,especially in a highly dynamic network environment.Otherwise,the failure of single component service may lead to cascade effect,which can make the whole system unable to work properly.Therefore,it is particularly important to study the quality assurance of component services for the whole service-oriented system.As a means of quality assurance of Service-Oriented Systems,reliability prediction has been studied by many experts and scholars in recent years.The present study aims to solve the problem of online reliability prediction of component services in Service-Oriented Systems under dynamic fluctuating network environment in large data environment.But it still faces the following four challenges: the parameters reflecting the reliability of component services are difficult to define and obtain;the variation of the reliability of component services in highly dynamic and fluctuating network environment does not have obvious regularity;in order to provide proactive guidance for Service-Oriented Systems,the prediction algorithm is better able to provide real-time reliability of component services for a period of time in the future;and how to improve the accuracy of the reliability prediction model of component services in large data environment.To address the above challenges,the present study carries out research and completes the following tasks:(1)define the parameters reflecting the reliability of component services firstly;(2)and then use clustering algorithm to find the characteristic values of parameter sequences to solve the problem that the original data set does not have obvious regularity;(3)and finally propose two models for the real-time reliability prediction of component services in large-scale data environments: the model based on the combination of Deep Belief Network and Long ShortTerm Memory Network and the improvement of the model: the combination of Deep Belief Network and Bi-Directional Long Short-Term Memory Network.Among them,Deep Belief Network can compress the features of high-dimensional data,and Long Short-Term Memory Network and Bi-Directional Long Short-Term Memory Network can effectively process temporal data.(4)In order to verify the availability of the two models,a series of experiments have been carried out to compare with the existing classical algorithms for reliability prediction.The results show that the proposed models have obvious advantages in the reliability prediction of component services.
Keywords/Search Tags:Service-Oriented Systems, Reliability Prediction, Deep Belief Network, Long Short-Term Memory Network, Bi-directional Long Short-Term Memory Network
PDF Full Text Request
Related items