Cloud computing is a concept of providing user and application oriented services in a virtual environment. Users can use the various cloud services as per their requirements dynamically. Different users have different requirements in terms of application reliability, performance and fault tolerance. Static and rigid fault tolerance policies provide a consistent degree of fault tolerance as well as overhead. In this research work we have proposed a method to implement dynamic fault tolerance considering customer requirements. The cloud users have been classified in to sub classes as per the fault tolerance requirements. Their jobs have also been classified into compute intensive and data intensive categories. The varying degree of fault tolerance has been applied consisting of replication and input buffer. From the simulation based experiments we have found that the proposed dynamic method performs better than the existing methods.