Until recently, almost no IT industry vendor or analyst questioned the assumption that nearly all kinds of virtualisation deliver quick, significant cost savings compared to computing only in the physical world.
Server virtualisation, in particular, creates such a quick payoff during the early stages of a P2V migration that it often gets a pass on strict ROI/cost justification as part of the approval process, according to Ian Song, research analyst in the System and Virtualisation Software group at IDC.
"People got a really good impression from server virtualisation because the concept is simple, and the payback is within something like six months, because you can put four servers or more into one physical box and save the cost of all that hardware" he says. "Desktop virtualisation, especially, but cloud and streaming apps and other models all tend to suffer by comparison."
Higher than expected costs or lower returns are major reasons for VM stall, the slowing or stoppage in large virtual server migrations, according to James Staten, principal analyst at Forrester Research.
Some cost surprises come from unrealistic expectations, some from the failure to adapt IT organisations or budgets to take greatest advantage of the new technology, Staten says. But some surprises also come from the tendency of even financially savvy top IT managers to avoid reporting specific costs of virtualisation or cloud projects to their bosses, according to a study funded by systems automation vendor Apptio for the Worldwide Executive Council.
The study found that 64 percent of IT managers believe detailed tracking and analysis of virtualisation costs will be important this year, and 20 percent thought it would be critical.
However, 48 percent of respondents say they report the cost/benefit of cloud or virtualisation projects as a single lump sum, the cost. Twenty five percent track the cost, but say their numbers aren't accurate enough to use for budgeting or audits. Twenty percent don't report cloud or virtualisation spending as separate items. Ten percent don't track the costs at all.
Eighty percent said more detailed reporting would be more important this year.
The biggest problem seems to be that most IT managers just don't know how they should measure or report the results of virtualisation projects, Staten says. They're able to get away with that because so few companies use chargeback to track IT spending by particular departments, let alone chargeback specifically for cloud or virtualisation, according to Galen Schreck, vice president and principal analyst at Forrester.
Some are using "showback" - the same thing as chargeback except IT only reports what the cost of a business unit's virtual IT use would have been, without actually charging for it.
Even given the willingness to do solid financial analyses, there is some debate about how, specifically, to charge for computing services abstracted from the hardware, storage and networks.
There are four basic approaches according to analyses and recommendations from Apptio, by Andi Mann, former analyst at Enterprise Management Associates, who is currently VP of virtualisation product marketing at CA Technologies, and by Gerod Carfantan, a solution development manager at VMware who blogs about cost and performance analysis at vCornerOffice.com:
This is essentially consumption-based costing, in which total IT costs are divided according to the volume of transactions, number of servers, number of users or other standard measures of the volume of IT's work as consumed by a particular business unit.
This method works if the analysis and tracking is clear even in a virtualised environment and there is a way to account for unused capacity. What business unit pays for virtual machines dedicated to applications that no one seems to be using? Pure cost per minute costing is attractive from an external service provider, but is only possible because other customers pay for that capacity when you're not using it.
Branch offices that don't need huge storage, high bandwidth or additional support could pay a lower "base rate" than larger, more resource intensive business units.
Those that handle most of their own support, can provision and manage their own servers and manage other tasks that would otherwise fall to IT might enjoy rebates or discounts on the base cost.
Service costs vs infrastructure costs
One way to account for unused capacity is to separate "base" costs like WAN bandwidth or data centre real estate from the applications, network access or other services a business unit uses. Base cost to the business unit would remain relatively stable, while service costs would vary according to consumption.
Virtual server density, the cost of all the infrastructure required to operate the virtual infrastructure, licences, servers, storage, bandwidth, IT salaries, real estate, utilities, etc., can be totalled and divided according to the number of virtual servers a business unit requires.
Or, costs can be divided according to the requirements of the application. All the same elements go into it as into virtual server density calculations, but costs are divided according to the resources required by a specific application. Costs to a business unit are determined according to whether it is the only one using that application, or according to its proportional use compared to other business units.
Direct or indirect weighting is a way of dividing costs according to an external calculation of the demands from a particular business unit. This can involve calculating a department's IT cost according to the percentage of the company's total operating budget required, for example, or total headcount.
Which method is right for you? The approach that works for a specific company will vary according to its particular culture and accounting structures, Staten says.
Getting the costs straight is more than just an accounting game, however.
Until business units know what IT resources they're using and have the ability to change usage based on their own needs, they won't see nearly the direct, tangible benefit of the flexible capacity of either virtualisation or cloud computing, according to Gary Chen, research analyst covering server virtualisation at IDC.