A couple of month ago, ESRI Inc. introduced its Amazon Arc GIS Server AMI. It is a convenient way to use ArcGIS Server in a Public Cloud Environment. However, Public Clouds such as Amazon Web Services have some drawbacks in terms of trust or responsibilities for lost data. In some settings it might be impossible to migrate to AWS due to these issues. Another aspect is the costs that are associated with AWS.
With 0.48$ per hour for one large windows instance (ca. 357$/full month) a significant amount of money can be accumulated for a system running 24/7 over a larger period of time.
Private Clouds address these issues. On the one hand, no data is given away to a third party and the owner of such a Private Cloud has still full control over the data. On the other hand, there are no direct costs associated with a Private Cloud (there are of course indirect costs, such as maintenance, utilities, hardware and software license costs etc.).
An ESRI ArcGIS Server AMI can unfortunately not being used in Private Clouds. This was the starting point for us to explore methods to fill this gap. We used our Private Cloud based on the Eucalyptus Framework with our own Cloud Management Software and deployed an ArcGIS Server in it (after several technical pitfalls) . It is now possible to route incoming traffic to different ArcGIS Server instance and scale the number of ArcGIS Server instance based on the amount of incoming requests. Next steps are to try different setups with split SOM and SOC and enhancing our scaling and load balancing mechanisms.