I have created a Hadoop cluster with 3 nodes. When the cluster is not in use (no hadoop or spark jobs running) , there are 3 ways. Keep the Virtual machines running all the time which will be and expensive option or shut down the Virtual Machines when not being used or to delete the virtual machines and create new machines and setup the cluster each time. The first option is the most expensive while the last one is cheapest one but difficult one. What is the recommended way?