When IT assets are owned, the rule of thumb is the higher the usage, the lower the unit cost per user. This is because the infrastructure costs can be amortised over a larger user base. Investing in IT for only a few hundred users can cost a lot more per head than for 1,000 users.
However, Meta Group analyst Rob Schafer said when usage is purchased on demand, that rule goes out the window. Typically, an on-demand, utility computing supplier will agree a price for a level of users, but charge a premium if that level is exceeded.
So when demand spikes over what has been anticipated, the cost per user will increase as more users come onstream. Even worse, said Schafer, if demand drops below the agreed level, the cost per user will also increase.
IT directors must understand that their greatest risk, and thus cost, is demand volatility, and it is there that most analysis must focus to determine the pattern of user demand. Get it wrong, fail to predict excess or under-demand, and it will cost you.
The question is, will it cost you more than it would to own, instead of rent, infrastructure capacity to meet such spikes? How much capacity would have to be built in if those spikes are not predictable?
Purchasing is insufficient capacity could lead service levels to fall below what is acceptable to business, but over-generous capacity is costly and wasteful. Paying a premium to an on-demand utility supplier could be cheaper.
Some spikes are predictable, such as month or year ends, but they still need to be allowed for in the contract. Similarly, said Schafer, IT directors must strive to predict general growth in demand as accurately as possible and build it into the contract.
These are the three uncertainties that go into the cost equation when deciding whether to buy on-demand IT: what is the growth in usage; how much does the peak demand differ from the base level; and how unpredictable is that demand?
This was first published in February 2004