Research On Fault Tolerance Strategy Of Incremental Checkpoint Based On Artificial Potential Field

The on-demand service characteristics of cloud computing make it applicable to large-scale scientific research scenarios as well as individual users,making it one of the most popular distributed technologies.Its dynamic service architecture and the complex characteristics of basic equipment make fault tolerance a problem that cannot be ignored in cloud computing.Because the failure of the cloud computing system will not only bring irreparable economic losses,but also lead to catastrophic consequences.Therefore,checkpoint technology,as the most commonly used fault-tolerant technology in cloud computing,has become a current research hotspot,and the fault-tolerant strategy based on incremental checkpoint is the main technical means of fault tolerance in cloud environments.The content of the research results obtained in this thesis is as follows:In view of the problem of excessive checkpoint recovery cost in the cloud environment,considering the checkpoint recovery cost and setting cost,the incremental checkpoint and the global checkpoint are combined to form an incremental checkpoint fault tolerance algorithm.At the same time,the checkpoint frequency setting also comprehensively considers: failure probability,task completion time,dynamically changing the checkpoint setting interval,and minimizing checkpoint overhead.Through experiments,it is verified that the algorithm in this thesis can not only target fault tolerance of different types of faults,but also effectively reduce the cost of system optimized fault tolerance,and provide a new solution for cloud fault tolerance strategy.In order to improve the reliability and availability of checkpoint files,this thesis designs an incremental checkpoint storage strategy based on artificial potential field for the problem of checkpoint layout.First,the proposed fault-tolerant model is used to obtain the checkpoint setting time series.Secondly,according to the analysis of node resources,the construction of artificial potential energy field is discussed.The dynamic change of global resources is used to achieve the matching of checkpoint files.Finally,it is enhanced by hash encryption algorithm and digital signature technology.Checkpoint file reliability and security.Experiments prove that the strategy proposed in this thesis has obvious advantages in terms of execution time overhead of storage tasks and load balancing of data nodes in the system.Finally,this thesis designs and implements a hierarchical checkpoint fault-tolerant system based on the Python Django framework.All the research results of this thesis are explicitly verified,which provides a new realistic solution for the fault-tolerant technology of cloud systems.