At the recent Structure 2010 "Year of the Cloud" conference, Amazon.com CTO Werner Vogels railed against "the false cloud"--the idea that cloud computing can be done internally, within an enterprise. The "true cloud" in this worldview can only be realized using the facilities and resources of network-accessible service provider. Anything else? False. A misappropriation of the cloud name and concept--a pretender that cannot deliver the benefits of cloud computing.
Oddly, another big theme of the conference was the idea that cloud is not defined by specific technologies, but rather by the benefits it delivers. These viewpoints could not be more starkly opposed. One one hand, you have to do cloud a certain way, or else it's a false pretender. The more relaxed "if it quacks like a duck..." view says if you get cloud-like features, benefits, and outcomes, your cloud is perfectly real enough.
That Vogels defines cloud as he does isn't surprising. His company's Amazon Web Services unit is the paradigm of the service providers he deems essential to the "true cloud." Many other cloud promoters similarly work for companies that sell external cloud services, or, one step removed, are aligned with those who do. Beyond vested interests, there are emotional ties. Cloud enthusiasts are frequently techno-optimists; they enjoy living on the leading edge. Apps run in an enterprise data center? How yesteryear! How old school! Rip it out and go this much better new way!
Personally, I love external cloud providers. Like many developers, Web operators, and small to medium businesses, I use services such as Amazon Web Services, Rackspace, Google, and others in preference to in-sourced IT. I recommend that others use such services as well, whenever they can. But "use them when and where they make sense" is the key.
What suits me, SMBs, developers, or Web companies isn't necessarily what suits large enterprises and their applications. Over the next few years, enterprise IT will not choose to run the majority of its IT estate within external cloud service providers. Even those pining for cloud's flexibility, elasticity, and efficiency would not be able to shift immediately. The reasons include things like organizational inertia that we might all bemoan--but also very real performance, capability, customizability, quality-of-service, security, and regulatory/auditing issues with today's cloud. However delightful and revolutionary, cloud remains in its early days.
If cloud computing is going to mean something practical, important, and central to enterprise IT over the next few years, cloud must be broad enough to include privately-owned and operated infrastructure. Cloud's benefits must be accessible to enterprise apps. That will often mean implementing the benefits with heavier performance, availability, and other guarantees than available in today's version 1.x cloud environments.
One of the most exciting opportunities is hybrid clouds--the combination of enterprises' internal/private facilities with external/public cloud resources. Hybrid clouds provide a great transitional strategy for enterprises to get started immediately and practically, then to evolve where and how apps are deployed over time as their requirements, service providers, and cloud comfort levels change. The hybrid/enterprise cloud will tend to be built atop virtualization tools, such as VMware's vCloud, rather that the lighter-weight, more bespoke tooling currently consumed by Web operators.
In the end, if cloud purists want to insist that only external clouds qualify, that's their right. But as cloud rapidly mainstreams, enterprises will demand that their concerns and approaches be taken seriously. For enterprises, cloud defined by its benefits is the true cloud. Hybrid extensions of virtualized infrastructure is the most likely path. Notions of cloud more narrowly defined--for example, constraining where the apps are run--those will be the false pretenders.