Mobile applications' requirements for compute capability grow up daily. Providing mobile service in remote cloud is one of the solutions to this issue. However, due to the long geographical distance of clouds, it is difficult to ensure the claimed performance of real-time service. As a result, mobile edge computing has become the main solution to solve this problem. The present researches are concerned with cloudlet placement, and computation offloading, while the problem about how to deploy services on cloudlets based on the user's geographical distribution is overlooked. We address this issue in our paper. With the consideration of minimising the number of services and guaranteeing access delay constraint at the same time, this problem is formulated into an optimisation problem. For the NP-hardness of the problem, an efficient heuristic algorithm is proposed to resolve it. The simulation experiment at the end of this paper evaluates the performance of this algorithm. Experiment results demonstrate that the proposed heuristic algorithm is effective.