In this work, we investigate factors that can impact the elasticity of bare-metal resources. We analyse data from a real bare-metal deployment system to build a deployment time model, then use it to determine how long it takes to deliver requested resources to cloud users. Simulation results show that reservations can help reduce the time to deliver a provisioned cluster to its customer, by enabling machines to be started in advance or be kept powered on when there are impending reservations. Such an approach, when compared to strategies that switch-off idle resources, shows that similar energy savings can be achieved with much smaller impact on the time to deliver the provisioned clusters.