A method of operating a resource management node includes, for each of a plurality of resources available in a plurality of distributed computing systems, determining a productive spend metric based on utilization of the resource and cost associated with making the resource available in the distributed computing system. Relocation of workload of selected ones of the resources between the plurality of distributed computing systems is controlled based on the productive spend metrics determined for the selected ones of the resources. Related computer program products and resource management nodes are disclosed.