The number of enterprises using cloud computing will grow to 9% in three years, says one analyst, but its impact will far exceed those modest figures.
Late last year, the market research group IDC surveyed IT professionals and concluded that 4% of enterprises already have implemented some form of cloud computing, although it's often in the form of software as a service (SaaS), such as Salesforce.com's CRM application.
That number will more than double by 2012, to 9% of enterprises, said Frank Gens, senior VP of IDC, as he opened the Cloud Computing Forum in San Francisco on Wednesday.
But the impact of cloud computing will far exceed those modest figures, he added. Applications designed to run in the cloud "will represent 25% of the net new growth in IT spending," versus spending for on-premises IT, he predicted. SaaS by itself is projected to nearly double from $9 billion to $17 billion in that time period, Gens said.
Another forum speaker, Joseph Tobolski, director of cloud computing at Accenture's Technology Labs, said in an interview that he thought the figures were too low.
Tobolski pointed out that if he worked at a company with no commitment to cloud computing, it would still be a participant because he backs up his laptop to the Amazon EC2 cloud every few days. If his laptop turns up missing, he will still have his data and documents available, he said.
Achieving a 5% to 9% adoption rate through 2012 "might be understating it," he said. It's a mixed blessing that cloud computing is so easy to start using in stealth mode. Business users seeking more computing resources don't have to ask IT for permission to swipe their credit cards and rent cloud services.
"That may be an underreported phenomenon," he said. If you ask Amazon to list how many users come from an easily recognized corporate address, then ask the CIO of the company how many cloud users he has, "it might be hard to match up those two answers."
Tobolski said Accenture's cloud consulting services are growing, and it's even produced an enterprise service bus to make it easier for its clients to link their data centers to cloud services. IT managers should be trying to make their internal operations function more as an enterprise cloud.
"Until you can get your internal stuff to act like an external cloud, you won't get the full benefit," such as moving workloads easily from the data center into the cloud, as needed, he said during his talk to about 250 attendees at the forum.
In an interview afterward, however, he warned that getting internal operations to resemble a cloud has more to do with rearchitecting the computing infrastructure to offer services. If the data center is siloed, with isolated applications running on single servers, then the goal of building an internal cloud will simply lead to another, albeit new form of, silo -- a set of isolated cloud servers.
Heavy reliance on external cloud services, without the ability to monitor how former internal business processes are running, could lead to a new IT ill, "cloud services sprawl," with it unclear which business process is using which cloud services.
Tobolski foresees prospective uses of cloud computing that he thinks are logical. One is as an aid to integration in mergers and acquisitions. If two companies have the same application, a new application can be constructed in the cloud that they can access from their respective data centers and use to gradually displace different systems.
He also thinks it will be possible to run legacy systems in the cloud, although few users of cloud computing do so today. If the cloud's efficiencies of scale become convincing enough, then emulation programs or virtualized versions of old Cobol, Fortran, and other legacy language applications will be moved to the cloud to take advantage of them, he predicted.
Most cloud computing is done in PHP, Python, C, C++, C#, or Java, languages more closely associated with current applications and the Web.
Accenture defines the cloud as "a collection of network hosted services available from outside the data center. The services are programmatically controlled (or automated to respond to demand). It often provides an illusion of infinite capacity" because the services can expand as needed, Tobolski said.
He admitted that with many vendors rapidly adopting the term, there's a lot of "hype connected with cloud computing," but don't be fooled: "There's something there in all this fluffy stuff."
To bolster his statement, Tobolski cited a definition from Wikipedia: "Dynamically scalable and often virtualized resources [that are] provided as a service over the Internet." That definition wouldn't satisfy everyone, but how could it be wrong? he asked. It's a definition that came out of the cloud, he noted wryly.
IT Service Management Must EvolveThe idea of technology being delivered as a service appeals to the 409 IT pros responding to our Service-Oriented IT Survey. But cloud providers are competing for that work, and CIOs are being selective.