Massively Multi-players Online Gaming (MMOG) presents one of the most popular large scale Cloud Gaming services. However, minimizing the overall MMOG operating cost is viewed as a critical task. The main challenge for MMOG providers is to find the best tradeoff between cost and response delay when allocating resources: in fact, allocating powerful resources lead to an excellent operating delay, but it is costly. However, cheapest resources don't guarantee even acceptable latency especially for high peak periods. In this view, the present paper contributes to optimize, under delay constraint, the resources allocation cost of MMOG service over a distributed Cloud infrastructure. We develop a model to capture the intrinsic trade-off between response delays relative to allocated resources and their corespondent costs. Then, we formulate the optimal VMs placement problem, which is NP-hard. Besides, we propose a Cost-aware Virtual Machines placement algorithm that approximates the best trade-off point between the two compromise terms. We evaluate our contribution comparing to a Random, Greedy Allocation Cost (GAC) and Minimum Allocation Cost (MAC) placement algorithms via Matlab tool. Results show elasticity and effectiveness of our contribution in maintaining the balance between cost and delay.