Cloud Computing refers to the use of memory and computing capabilities of computers and servers around the world, so the user has considerable computing power without the need for powerful machines. The probability of failure occurring during the execution becomes stronger when the number of nodes increases; since it is impossible to fully prevent failures, one solution is to implement fault tolerance mechanisms. In this work, we have developed a fault tolerant service based on the checkpointing in cloud computing. Our fault tolerance service uses a semi-coordinated checkpointing that minimizes the consumed energy and the overhead by decreasing the time of the coordination phase. Our service also decreases the rollback cost. The experimental results show the effectiveness of our proposition in term of execution time, energy consumption of SLA violation.