Cloudonomics: The Business Value of Cloud Computing, + Website

If Clouds Are More Expensive

Suppose that the cloud is more expensive on a unit-cost basis. The key then becomes the magnitude of the peak-to-average ratio compared to the utility premium. If the demand is spikier than the cloud is relatively expensive, the cloud still can be cheaper overall. Specifically, suppose that U < P/A. By multiplying both sides by A, we see that A × U < P. Multiplying both sides as usual by cr × T, it turns out that A × U × cr × T < P × cr × T, and again the cloud solution is less expensive. Suppose that it costs $200 a day to rent a car. If you need the car only for one day a month, it still makes sense. The peak requirement is one car, but the average demand is only 1/30. Therefore, the peak-to-average ratio is 30, whereas the utility premium, even at that exorbitant rate, is only 20. Since the peak-to-average ratio is larger than the utility premium, the utility pricing model of the cloud is preferable. Conversely, if the demand is flat, at least compared to the utility premium (i.e., where U > P/A), a dedicated solution would be less expensive than the cloud. Then A × U × cr × T > P × cr × T. However, under such circumstances, it turns out that a hybrid solution of both dedicated and on-demand resources is optimal, as we’ll explore next.