Enterprises that migrate to the cloud assume the path they choose -- whether private or public -- will be cost-effective in the long term.
That isn't always the case.
An enterprise that doesn't carefully weigh its consumption model prior to a cloud migration could set itself back and overpay.
There are two types of consumption models in an enterprise: fixed and variable. A fixed consumption model includes workloads that use the same amount of resources at all times. A variable model involves workloads that frequently increase or decrease the amount of resources it needs to perform.
The consumption model should be part of the formula an enterprise uses to determine what type of cloud to install.
David S. Linthicum,
vice president of Cloud Technology Partners
For example, enterprise databases probably have a fixed consumption model, while Web applications are often variable and better suited in an elastic environment.
Problems arise when IT pros don't factor in the amount of resources an application uses before they move to the cloud.
"Cloud is often associated with people trying to save money," said Pascal Vitoux, CTO at ASG Software Solutions, an enterprise management company based in Naples, Florida. "When customers go to the cloud in rush mode, they often fail to realize there is a ton of investment they weren't expecting."
A public cloud could actually cost more in the long run than an in-house private cloud build for companies with fixed consumption models. Certain expenses come into play, such as rising costs of network connectivity required to manage an application in a public cloud, regardless of how low Amazon Web Services, Windows Azure and Google cut their prices.
"[Many companies] do a calculation on a fixed consumption [and find] the cost of moving to a public cloud is the same as private after two or three years," Vitoux said.
Enterprises don't understand cloud cost models as well as they should, according to David Linthicum, vice president of Cloud Technology Partners in Boston.
"You have to build fairly complex and wide-reaching total-cost-of-ownership models," he said. "The public cloud is not always cheaper, particularly if you've already made an investment in hardware and software."
Conversely, an organization with unpredictable resource usage can end up with overprovisioned and underutilized applications.
But a public cloud could scale those apps on demand to accommodate spikes in use.
An enterprise must also look at business factors such as workload profiles and security and management controls that coincide with its consumption model.
"The consumption model should be part of the formula an enterprise uses to determine what type of cloud to install," Linthicum said.
A logical evolution to private cloud
For some companies, the decision to migrate to the cloud is not a conscious one, but a necessary business move after a data center consolidation.
The Yellow Pages Group (YPG) in Montreal, Quebec couldn't run everything in a public cloud due to security and government regulations, so a private cloud was the right choice from a business and consumption standpoint.
"Since we had to have a bunch of servers, it made sense to virtualize them," said Alain Gaeremynck, senior enterprise architect at YPG. "Then we had this realization that virtualization alone wouldn't allow the level of flexibility needed to move as fast as we have to, hence the private cloud.
"It's sort of a logical evolution from metal to virtualization to cloud," Gaeremynck said.
While the private cloud made sense for YPG, the challenge for smaller to midsize enterprises is to stay on top of the constantly evolving technology -- whether it has a fixed consumption model or not.
"That requires a very knowledgeable staff and that can be fairly expensive," Gaeremynck said. "Building a private cloud infrastructure takes time and it's a huge investment. So even if you can predict with precision your consumption and growth, it might still be cheaper to go public."