With the rapid development of mobile cloud technology,the complexity of mobile cloud system is increased,and the probability of system failures is rised dramatically.These failures not only cause the huge economic loss of service providers and users,but can also lead to serious catastrophic events.Therefore,both the research community and the government-user agencies attach great importance to the study of fault-tolerance on mobile cloud.Whereas,fault-tolerance strategies based on hierarchical checkpoint technology have become the crux technique in the current mobile cloud fault-tolerance field.The thesis mainly focuses the technology of hierarchical checkpoint.The primary achievements of the thesis include as following:1.In view of the fault-tolerance overhead of mobile cloud system.According to a random renewal reward theory,integrating with the probability density function(PDF)and the cumulative distribution function(CDF)of the system,a fault tolerance algorithm based on hierarchical checkpoint is proposed to optimize the large fault tolerance cost of mobile cloud and dynamically determines the frequency of the hierarchical checkpoint.The numerical experiments prove the algorithm not only can cope with diffident species failures of the system,but also optimize the cost of fault tolerance,which provides a new solution for fault tolerance in mobile cloud.2.In order to improve the service quality of mobile cloud systems,hierarchical checkpoints must be high availability and high reliability.Therefore,a hierarchical checkpoint storage strategy based on supergraph coverage is designed in this thesis.Firstly,the fault tolerant algorithm proposed in this thesis is used to obtain the time series of hierarchical checkpoints.Then,based on hypergraph theory,the storage location of the hierarchical checkpoint is decided.Experiments show that the proposed strategy has obvious advantages in the execution time overhead of storage tasks and load balance of data nodes in the system.Finally,a hierarchical checkpoint fault-tolerant system based on the Python Django framework is designed and implemented.The system can explicitly verify all the research results in this thesis and provide a new realistic solution for the fault-tolerant technology of mobile cloud system. |