Total cost of utility
- Reduce text size Decrease text size
- Increase text size Increase text size
- Print article Print
- Jump to comments Comment
- Share this article Share
- Email article to a friend Email
Getting the best value from utility computing may be more complex than first thought.
As most CIOs will acknowledge, pressure is still being exerted on IT departments to deliver more value from their budgets, but they are being allocated little in the way of additional funds to drive this extra value. It is therefore reasonably sound to conclude that any move towards grid and utility type infrastructures must be based on sound financial arguments for doing so.
Industry heavyweights such as Hewlett-Packard (HP), Sun and IBM have all identified grid computing as being the next disruptive technology, albeit with subtly different visions as to the form it will take. At the heart of the marketing hype that has accompanied their evangelical passion for grid computing lies a nugget of truth: current server utilisation is woefully low. Businesses are spending money on servers whose processing power is used only intermittently.
But while the desire to drive value from theIT infrastructure might sway some companies towards investing in internal grid capabilities, a quicker switch to outsourced utility computing may tempt others. Here, vendors such as Sun Microsystems and Savvis have made initial steps towards offering an entirely outsourced IT infrastructure package, where customers only pay for the processing power they consume.
Clearly, senior IT management needs to decide which of the two approaches might best suit their own organisations.
Grid costs
A common perception of enterprise grids is that they will be built from cost-efficient, commodity hardware, and the popularity of blade servers has done much to reinforce the view. This theory suggests that businesses can attain mainframe-level power by building a virtualised network of these blade servers at a much lower cost.
Certainly, the headline figures have been impressive: IBM has been developing its Virtualisation Engine for its Power chip-based servers, estimating that a virtual array of these servers could cost as much as $30,000 less than an equivalent mainframe system. Some grid proponents go as far as to suggest that grids will also be built with self-tuning features that can drastically reduce management costs.
But how realistic is the grid computing nirvana? Virtualisation has been available in the mainframe world for a number of years, but it is a relatively immature concept for commodity servers. "Before firms achieve integrated management that spans platforms from multiple vendors, they must first rationalise the hairy management mess that they've cobbled together over the last decade," says Galen Schrek, an analyst at Forrester Research. This is not a task that will be completed overnight.
| ||
IT managers will also have to be mindful of software costs associated with building grid architectures. Most operating system and application licensing policies require one licence per instance; under those terms, a company could face having to buy four software licences if it operates four virtual servers.
Utility as a service
The notion of utility computing delivering 'IT on demand' has led some to consider it as an outsourced service, and already, some vendors are gearing up to deliver this. "Switching from a capital to operating expenditure model can be a major financial benefit. Service providers that have the necessary technical and process skills can offer compelling services to organisations looking to make the most of the IT infrastructure they buy," says James Eibisch, hosting services director at analyst group Gartner.
Managed services company Savvis launched its hosted utility service offering in autumn 2004, promising to slash the costs of IT. Savvis claims it will be able to cut customers' IT costs by 40% or more, as a result of the economies of scale offered by sharing integrated virtualised computing, network-switching products and storage resources among customers. "It offers a real chance to lower cost of ownership of infrastructure," says Rob McCormick, CEO of Savvis.
This is an area where hardware giant Sun Microsystems has also been keen to establish a user base. Sun expects to launch its Secure N1 Grid service, which delivers 'pay per use' services over the Internet, by the end of 2004.
Initially, it will offer the service at $1 per CPU hour and 80¢ per gigabit of storage. Speaking at Gartner's US IT symposium, Sun CEO Scott McNealy said that the offer could even damage Sun's traditional server business: "I'd rather we cannibalise ourselves rather than have someone else do it."
But while these initial offerings may look tempting, pricing models still lack sophistication, says Gavin Williams, infrastructure director at systems integrator Avanade. "There's no IT equivalent definition of the kilowatt hour. Do you need to be pricing for CPU power or number of transactions? They're both very different requirements," he says.
With both internal grids and outsourced utility options offering the potential to reduce costs, CIOs need to take a long term view of their IT strategy, says Rakesh Kumar, of analyst company the Meta Group. "IT executives will need to stop viewing consolidation projects as revolutionary cost-cutting activities and instead approach them with a process-driven, continuous change management strategy." This will allow them to amalgamate the steps towards utility computing with their existing environment, he says.
| ||||||
