McKinsey & Company says that outsourcing your typical big corporate or government data center to a cloud service would cost double what it costs to keep the thing in-house.

The mighty consultant has run up what it figures is a "hype-free" study called "Clearing the Air on Cloud Computing" that recalls the dot.com bubble and warns of the dangers of investing in hype.

Its conclusion is that "cloud computing has shown great promise for start-ups and pet projects for large corporations. However, cloud computing can divert IT department's attention from technologies that actually deliver sizeable benefits."

McKinsey used Amazon's pricing as its point of comparison and determined that the total cost of the data center's functions would be $366 a month per unit of computing output while keeping the work at home only costs $150 a month.

And as for the CAPEX-saving come-on of the cloud, well, owning hardware is more cost-effective than renting it if you throw in the tax deductions for depreciation.

Labor costs are pretty much the same no matter what you do. McKinsey finds 10%-15% of your headcount can be cut by migrating capacity to the cloud, most of it coming out of facilities and touch labor roles like administrators, but the savings is offset by the difference in hardware pricing.

Then there are the abiding security and reliability issues of cloud, the fact that applications may have to be re-architected and that business perceptions of increased IT flexibility and effectiveness will have to be managed. McKinsey also notes that enterprises tend to set their SLA uptimes at 99.99%, a figure cloud providers aren't ready to match.

It says the cost of the cloud has to come down about 144% to make outsourcing a complete data center economically sensible.

It allows that the cloud may make sense for some small and medium-sized companies with revenues of no more than $500 million - if they limit themselves to pre-paid Linux servers.

It advises big enterprises to use virtualization, the more aggressive the better. It reckons they can get server utilization rates similar to cloud providers.

It figures going in that a typical big-time data center has 10% capacity utilization. That number, it claims, can be increased to 18% or even 35%.

And best practices, it says, can drive down server TCO by more than 50%.

The study includes a definition of cloud computing and says the publicly announced private clouds aren't clouds at all because the enterprise can't avoid capital expenditures. It says they're essentially aggressive virtualization on top of traditional enterprise IT stacks.

McKinsey defines a cloud as hardware-based services offering compute, network and storage capacity where hardware management is highly abstracted from the buyer, buyers incur infrastructure costs as variable OPEX, and infrastructure cost is highly elastic (up and down).

See here.

• McKinsey & Co - Clearing the Air on Cloud Computing
• McKinsey Report Aims To Keep Expectations About Cloud Computing Realistic
• Cloud Computing Debate: Booz Allen Hamilton Comments on Recent McKinsey & Co. Report