The combination of on-premise and public resources into one compute pool is calle "Hybrid Cloud". The idea behind is to use a public pool to extend the local resources.
This is much more cost efficient than always planning for peak performance on premise. Our research - as well as our projects - show depending from the use case between 20 and 60% of operational savings.
But how does this look like?
Above diagram outlines a scematic strcucture of a hybrid setup:on the left the on-premise systems are shown on the right of the cloud the public resources. The interconnect between the two could be simply through the internet, channeled through a VPN or even created on a private line using an Interconnect, if available at your datacentre location.
The management of the remote resources will be done through either the cloud providers UI or - in the dynamic configuration - programmatically through a kind of Scale-out-manager. The latter might be triggered by events.
Triggering events could either be the passing of a threshold in your system load or a scheduled event. The Scale-out-manager will understand the event and process a configuration, which could then launch several machines in a public cloud, connect these to the particular network segment and include them in the local routing.
Thus will allow to decrease CAPEX and improve flexibility and even costs by reducing the amount of resources required for your operations to meet all kind of varying demand pattern.
The infratsructure required is fairly cheap as you will not need more than the local Scale-out-manager and triggering events. Once setup, this kind of solution will reduce your operating expenses drastically. However, pre-requisite to arrange such a scenario is the clear understanding of our demand pattern. This demand pattern you may simply retrieve by monitoring the behaviour of your application(s) using the ASCAMSO satellites.