A RESOURCE LEASING POLICY FOR ON-DEMAND COMPUTING.

The article presents a formulation and a solution method for the leasing problem for on-demand computing. The purpose is to minimize the cost of leasing resources while still keeping an adequate quality of service that we evaluate by the average wait time of requests. The problem is demonstrated as a infinite-horizon, continuous-time Markov decision problem. Results showed that the shape of the optimal cost function is convex-like in the number of resources leased and non-decreasing in the number of requests in the system. It is then concluded that when demand for the service is recognized to exist, but is unpredictable, it is advisable to lease more than fewer resources.