The future of corporate IT is in private clouds, flexible computing networks modeled after public providers such as Google and Amazon yet built and managed internally for each business's users, the analyst firm Gartner says.
Cloud computing hype centers largely around the outsourcing of IT needs to cloud services available over the Internet. While this trend is expected to accelerate, Gartner predicts it will also become standard for large companies to build their own highly automated private cloud networks in which all resources can be managed from a single point and assigned to applications or services as needed. "Our belief is the future of internal IT is very much a private cloud," says Gartner analyst Thomas Bittman. "Our clients want to know 'what is Google's secret? What is Microsoft's secret?' There is huge interest in being able to get learnings from the cloud."
Bittman discussed Gartner's predictions in an interview with Network World, and will detail them again next month at the analyst firm's annual Data Center Conference in Las Vegas in a presentation titled "The Future of Infrastructure and Operations: The Engine of Cloud Computing."
While Bittman says it will take years for private clouds to develop, some early adopters are already "Google-izing" their own data centers. Bechtel, for instance, is using the software-as-a-service computing model internally to provide IT services to 30,000 users, in a project that relies heavily on server and storage virtualization. (Compare storage products.)
virtualization is key to building internal as well as external clouds, Bittman says, noting that Amazon hosts applications in Xen virtual machines. But server virtualization is only one of several necessary layers.
A meta operating
system -- similar to VMware's recently developed Virtual Datacenter Operating System -- will be necessary to manage an enterprise's distributed resources as one computing pool, Bittman adds.
Specifically, the meta operating system is "a virtualization layer between applications and distributed computing resources ... that utilizes distributed computing resources to perform scheduling, loading, initiating, supervising applications and error handling."
But the meta operating system only provides the muscles of a distributed environment, Bittman says. Another layer, which Gartner calls a service governor, will have to provide the brains, making decisions about where to allocate computing resources.
Say you have five business units and 100 applications -- some need ultra-fast performance and others don't. The service governor will decide which application gets what.
The technology "is evolving," Bittman says."This is not something that is just going to turn on."
Private clouds will take shape over the next few years, but perhaps only in large enterprises."Over time, a small business will not have economies of scale to make it worth staying in the IT business," Bittman says. Within five years, a huge percentage of small businesses will get most of their computing resources from external cloud providers, he predicts.
That's not to say enterprises with their own private clouds will shun cloud offerings that provide instant access to processing power and storage. Each company will manage a fixed capacity in-house and have access to external capacity from public providers when they need it -- sort of like overdraft protection in a bank account, Bittman says.
If a company experiences a sudden spike in demand, the meta operating system and service governor will arrange for extra capacity to be secured from outside sources. Users won't have any idea which server they are using or whether computing capacity is coming from inside or outside the enterprise, Bittman says. Layers of abstraction will permeate through the data center and users will be presented only with a services-oriented interface.
As public clouds evolve, enterprises will have many more choices, Bittman says. Today's cloud computing services boast of elasticity, the ability to scale resources up and down as needed at any time. Amazon's cloud computing service is called the Elastic Compute Cloud (EC2), for example.
But these services aren't truly elastic, in Bittman's view. Amazon charges a certain price for each virtual server and "if I want a slightly larger server, I have to buy another one. ... I have to do it in chunks,"