Techniques are provided for providing computing resources from a pool a plurality of networked computing systems to a consumer. The method includes determining that the consumer's resource usage exceeds a predetermined threshold. After a predetermined period of time, and upon determining that the consumer's resource usage continues to exceed the predetermined threshold, the method identifies one or more computing systems from the pool having capacity to host at least part of the amount of excess resource usage. The method further includes allocating resources on one or more computing systems selected from the identified computing systems to satisfy the amount of excess resource usage, and transferring at least the amount of excess resource usage to the selected one or more computing systems.