Even with the new support, you can’t just “plug in a server” and expect to use it for cloud computing, says Reuven Cohen, founder of the cloud-computing platform company Enomaly and the Cloud Computing Interoperability Forum. Instead, cloud computing relies on a series of layers. At the bottom is the physical hardware that actually handles storage and processing–real servers crammed into a data center, mounted in rack upon rack. Although companies are loath to disclose the size of their data centers, John Engates, CTO of Rackspace, says that hosting companies typically build them out in modules of 30,000 to 50,000 square feet at a time. Running on the hardware is the virtualization layer, which allows a single powerful server to host many virtual servers, each of which can operate independently of the others. Customers can change configurations or add more virtual servers in response to events such as increases in Web traffic. (It should be noted that not every cloud provider uses virtual servers; some combine the resources of physical computers by other means.)
Then comes the management layer. In place of platoons of system administrators, this layer distributes physical resources where they’re needed, and returns them to the pool when they’re no longer in use. It keeps a watchful eye on how applications are behaving and what resources they’re using, and it keeps data secure. The management layer also allows cloud companies to bill users on a true pay-as-you-go basis, rather than requiring them to lease computing resources in advance for fixed periods of time. Better billing may seem like a small detail, but it has turned out to be a key advantage over earlier attempts to create on-demand computing.