Virtualisation technology is the current darling of the software business, data centre managers and financial directors alike. Everyone likes its ease of deployment, maintenance, and usage -- but above all, they like the money it saves. The problem is though, how to quantify how much you've just saved.
Doing more with less has been the mantra of most businesses of any size ever since the Y2K debacle. Having to replace all those systems, sometimes from top to bottom simply because it was impossible to know whether or when a Y2K bug might strike, was hugely expensive.
The Y2K Cassandras were by and large heeded by enterprises, despite dissenting voices -- including that of the labs-test based magazine on which this author worked at the time, which argued that little evidence could be found that the problem would in practice generate big issues.
But what resulted was the adoption by most enterprises of a solid, business-based attitude to IT that negated the snake-oil charms of salesmen promising technologies and fixes for coping with the humungous burst of online traffic that, for many, never came.
As a result, IT managers have had to work a lot harder at getting more out of the technology they already possessed rather than fixing under-capacity by bolting new, untried stuff onto existing kit.
Virtualisation is key
Virtualisation is proving to be a key technology in that drive to extract maximum benefit from today's pretty potent hardware. Stories abound of organisations in both private and public sectors who, through virtualisation, have been able to cut the number of physical servers by a factor of five or more. Consequently, remaining systems work nearer their full capacity.
But a problem remains. It's not so much that an organisation needs to cost-justify the purchase of the software -- though that undoubtedly helps. It's more how to charge the cost of providing the service, now underpinned by virtual rather than physical machines, back to users.
With a physical server it's relatively easy -- and was probably yet another reason for installing each application onto a separate server in the first place. Add up the cost of the hardware and software licences, plus an element for admin, and other overheads, and you're done.
Virtual machines are another matter. Once an ESX server host is full of VMs and starts to run out of capacity, how can an IT department ensure that it's gathered enough money from its customers to add capacity by buying additional hosts and storage? If everything is shared, how do you allocate costs when some users gobble as much computing power you can give them, while others sip delicately.
There are a number of answers to this. As virtualisation market leader VMware puts it: "Virtualisation creates a variable allocation of resources that calls for a similarly responsive cost structure." The company reckons its charge-back solutions "allocate IT costs to business units based on variable resource usage in a virtual environment".
In other words, because of VMware's open Web services API policy, third parties such as Evident Software have developed VM-aware charge-back systems for use in production environments.
Evident is not alone of course. IBM's acquired privately-held software firm CIMS Lab in January 2006 for its virtual machine management software. The company's software allows companies to calculate the consumption of resources across an infrastructure that's likely to consist of a number of heterogeneous and virtualised platforms and systems. It's a move that Forrester Research analyst Frank Gillett says will change the situation. And IBM plans eventually to ship CIMS Lab with IBM Director, a suite of systems management tools that complements its xSeries, eSeries and pSeries servers.
One nuance of this issue is the need to charge users differential rates, allowing IT departments to create a number of tiers or levels of service, since not everyone needs the biggest, fastest server. The right product will eventually allow you to do this -- although today's generation of VM management systems doesn't yet support it, as BMC Software's David Wagner acknowledged. "Our product can give you resource utilisation by application across physical and virtual environments, but we don't put a dollar value on that," he said.
Why has this issue not been raised earlier? The answer is that it has, by many organisations. It's also true however that, in practice, the savings from virtualisation are often so compelling that it's worth making the shift to a VM world and worry about issues such as the charge-back mechanism later.
In the meantime, one solution that's making many IT managers happy, according to members of VMware's technology discussion forum, is a spreadsheet developed by Ron Oglesby , the author of a number of books and an infrastructure architect for RapidApp, in Chicago. Oglesby's specialities include large Citrix farm design, Intel virtualisation projects, and Windows AD design. Oglesby's spreadsheet, available from his website, is designed to help IT managers calculate how much a VM costs, based on individual environments -- although calculated in US dollars, it should be easily adaptable.
For many, however, there would appear to be but two options: wait for the right full-fat product to come along -- maybe it'll be IBM's -- or simply forget the whole idea.