Supporting this statement is Dennis Quan, CTO of the IBM High Performance On Demand Solutions. "We've designed the cloud around virtualisation. You have a datacentre with many servers and they are all turned into virtual machines," he says.

One difference from the now familiar multi-tenant SaaS model, in which numerous clients access a provider's application is that cloud computing environments also allow the customer to run his own applications on the provider's infrastructure.

According to William Fellos, an analyst at The 451 Group, the goal at a provider level, is to dynamically assign computing workloads as customer jobs come in. The approach helps the vendor maximise its resources and lets the customer ask for more computing power, on the fly. This is a key point, since a major goal of cloud computing, whether the IBM Blue Cloud or the Amazon EC2 (Elastic Cloud Computing), is rapid scalability.

But elasticity is probably a better term, says Barney Pell, founder and CTO of Powerset, a US-based start-up company building a natural language search engine. By elastic, Pell means the ability to stretch out when needed -- and then snap back. His company is attempting to index an enormous chunk of the web, a compute-intensive task that goes on most of the time. The work involves major spikes by users that would exceed the company's normal computing capacity.

Rather than buy enough servers and other infrastructure to meet peak needs, Powerset became an early customer of the Amazon EC2 and S3, the Amazon related storage service. "Powerset pays for the resources as it uses them, freeing up significant amounts of cash," Pell says.

Pell suggests that IT executives considering cloud services start by closely examining which resources their data centre uses all the time -- and which are only needed during periods of peak demand. What's more, the use of an elastic service gives IT time to establish a baseline, that is, the minimum level of resources needed to run the business at all times.

Similarly, groups or departments within enterprises often have the need to prototype or handle a specific project, but don't have the budget or desire to buy the required infrastructure. "IBM itself is using its internal cloud to supply the resources needed for prototyping new applications or services," says Quan. "Not every project uses that internal cloud, but more than 100 have," he adds.

Instead of buying hardware for the project The New York Times, for example, used Amazon Web Services (EC2 and S3) to generate PDFs of 11 million articles in the paper's archives in less than 24 hours using 100 instances of EC2, Derek Gottfrid senior software architect for the Times, wrote in his blog.

Flexibility up, costs down

For some enterprises, cloud computing can help a CIO tackle several problems at once, as was the case for Schumacher Group CIO Doug Menefee. Upon joining the Lafayette, US-based company three years ago, Menefee had to tackle a disaster planning gap and find new ways for IT to keep up with rapid business growth.

Headquartered two hours to the west of New Orleans and 35 miles north of the Gulf of Mexico, Schumacher staffs emergency rooms for 150 hospitals across the US. It only takes a glance at the map to see how close it came to being hit by hurricanes Katrina and Rita. "It was an eye opener," says Menefee. "We didn't have disaster recovery and business continuity capabilities. Had our headquarters gone down, it would have taken all of the regional offices down with it."