In this paper, the problem of commitment in allocating services to different applications and business processes is introduced. We model the problem based on the assumption that the services? instances are not put on hold throughout the lifetime of an application or a business process. Our objective is to find an optimal policy for committing services? instances to different applications to maximize the utilization of available instances of services. We formulate this problem and propose a Markov Decision Process approach for it. We present the optimal solution for a sample case with two types of services and two classes of application, and compare the performance of this optimal policy with a system with a Full Commitment Policy as well as a No Commitment Policy system. The comparison results show that the policy obtained outperforms the other two policies. We also evaluate the performance of a system considering beta distribution for the service execution time, and we illustrate the effectiveness of applying the obtained policy on this system as well.