"If you build it, they will come" may be a line from a baseball fantasy, but it describes what really happened in the world of cloud computing. One late summer day in 2006, online bookseller Amazon.com quietly made a portion of its excess data center capacity available to curious developers over the Web. Within hours, hundreds of developers had jumped at the chance to spin up some servers simply by opening their browsers and typing in their credit card numbers. Today, Amazon's Elastic Compute Cloud is the foundation for countless tech startups, and Amazon counts its business customers in the hundreds of thousands.
If you're a CIO with a data center to run, that has to make you think. Assuming you're not ready to entrust your company's crown jewels to the public cloud, couldn't you satisfy IT requests faster internally by borrowing from the Amazon playbook? Standardize on commodity servers and storage. Connect them using a converged IP network infrastructure. Layer on virtualization to support resource pooling, isolated multitenancy, automated provisioning, and dynamic scaling and contraction. Provide a Web front end and workflow engine that allows your "customers" to serve themselves, and bill them each month for what they use.
Unlike Amazon, you don't need to develop your own cloud management software. VMware, Microsoft, the OpenStack open source project, and many others are building cloud-oriented automation and controls on top of today's virtualization platforms. Depending on the number and sophistication of your users, you may not need or want elaborate self-service and chargeback mechanisms. Your on-demand resource pool might exist only to serve other IT groups within the company, as a shared platform for batch processing or short-term projects or to allow your development and test teams to do their thing without burdening server, storage, and network admins.