Research On Adaption Method Of Cloud Fault Tolerance Services Based On User Requirement And Resource Constriction

Characteristics of visualization and on-demand service make cloud computing attract many application systems migrating to the cloud.The complexity of cloud infrastructure makes cloud datacenters undergo many faults,and the reliability of cloud applications is reduced.So fault tolerance is important for cloud applications.When providing fault tolerance service to cloud applications,if cloud infrastructure providers provide services to cloud application,fault tolerance requirements of applications will not be considered,sufficiently.On the contrary,if applications implement fault tolerance mechanisms independently,resources of cloud infrastructures will not be well used.Therefore,the scheme of fault tolerance as a service is proposed,that is,the third party fault tolerance service providers provide fault tolerance services to cloud application in the manner of fault tolerant as a service.It can not only consider different application requirements for fault tolerance,but also make full use of cloud resources.In the related studies of fault tolerance as a service,most of them assume that users specify fault tolerance methods in their fault tolerance requirements.But the specification of fault tolerance method can lead to the following problems.Firstly,the main aim of fault tolerance is to improve the reliability and reliability is also the most concerned problem of users.The specification of fault tolerance method cannot reflect reliability concerned by the user directly.Next,if we let users specify fault tolerance methods,the required resources used to satisfy the requirements will be fixed.The stationary of the existing resources of providers and the stationary of required resources used to satisfy the requirements of fault tolerances will make providers provide fault tolerance services for a certain number of users.This fact makes resources cannot be used fully,and the income of the cloud fault tolerance service provider can be influenced.Therefore,in order to tackle the two problems above,i.e.,the description of fault tolerance requirements cannot actually reflect the reliability of fault tolerance and the problem that cloud infrastructure resources cannot be used fully,we propose a cloud fault tolerance services adaption method based on fault tolerant requirement and resource constriction.For one thing,we define fault tolerance requirements of cloud applications from the reliability of application components.For another,we study the fault tolerance services adaption method from the perspective of cloud fault tolerance services providers under the condition of resources of providers is insufficient or not and whether taking overhead generated by switching between fault tolerance methods into consideration.We generate fault tolerance services by using optimization theory.In this thesis,we give the reasonable description of the requirements of users,and calculate the required resources,overhead,response time when fault tolerance requirements are satisfied under different fault tolerance methods.Then we illustrate the effect of fault tolerance method,calculate the switching overhead generated by switching between fault tolerance methods,propose a fault tolerance service adaption method and generate corresponding fault tolerance services under different scenarios.The experiment results show that our fault tolerance service adaption method cannot only satisfy the requirements of users,but also make cloud fault tolerance service providers gain maximum benefits by making full use of resources.