Research On Fault Injection Model In Disaster Backup And Recovery System Based On Traffic Analysis

With the development of information technology, the demand for reliability in information system increases. Disaster backup and recovery system is used to guarantee information system free from the data loss, system halt and service interruption. In the study of disaster backup and recovery system, three major issues exist. Foremost, lack of traffic model aiming at disaster backup and recovery system while normal model cannot match the feature of this system; Second, current fault injection model is too simple and almost static, not considering the influence by fault correlation and system status; Finally, lack the study of trustworthy management of disaster backup and recovery system. Therefore, a fault injection model in disaster backup and recovery system based on traffic analysis is proposed after three aspects above, the main achievement is as follows:1. Propose a framework focusing on disaster backup and recovery system based on traffic analysis and illustrate every module’s function. The framework consists production center, disaster preparation center, network center, management center and audit center. The function of each part and the interactive mode between them are proposed and the simulation method of each module is established.2. Propose a traffic module in allusion to disaster backup and recovery simulation system and a traffic generation algorithm. The algorithm generates basic traffic flow with self-similarity and long range dependence using Pareto distribution in particular case. Also propose the conversion algorithm in allusion to normal business, special business and disaster backup and recovery business based on basic traffic flow which can describe current business system precisely. Several experiments verify the self-similarity of traffic flow and the availability of each conversion algorithm.3. Propose a fault injection model in allusion to disaster backup and recovery simulation system and a fault evolution algorithm. The algorithm uses Markov Process to analyze the effect on fault probability from system status and the change of traffic, solving the issue that little research on fault injection model which leads to static model usually. Experiments show that the model can simulate the fault injection process effectively and the correlation among faults through dynamic fault injection.4. Propose a trustworthy management model in allusion to disaster backup and recovery simulation system and develop a fast initialization algorithm of trustworthy rate and a fault isolation algorithm. Fast initialization algorithm of trustworthy rate build a connection between fault probability and trustworthy rate so that the trustworthy rate of new node can be acquired quickly. Fault isolation algorithm ensures the anomaly traffic flow can be limited in a small district while the system is still available with fault spread.