Cloudonomics: The Business Value of Cloud Computing, + Website

When Clouds Cost Less or the Same

If U < 1—that is, clouds are less expensive per unit of resource per unit oftime—the cloud is always a good idea, regardless of the application profile. This is easy to see. The cloud cost is A × U × cr × T, and the dedicated cost is P × cr × T. Since A ≤ P, by definition, and U < 1, we know that A × U < P × 1. Multiplying both sides by cr × T gives us A × U × cr × T < P × cr × T; in other words, the cost of the cloud is less than the cost in the enterprise.

If clouds cost the same, U = 1. The relative total cost then depends on whether A = P (i.e., the demand is flat) or A < P (i.e., there is some variation). There always is some variation, but let’s assume there is not. Then A = P, and since U = 1, we know that A × U = P × 1. Therefore, again by multiplying both sides by cr × T, we see that A × U × cr × T = P × cr × T. In other words, the cost of the cloud and enterprise approaches are identical. This makes sense. If the cloud costs the same on a unit-cost basis as dedicated resources, and the flatness of the demand means that we aren’t really using the special pay-per-use benefit, both approaches cost the same.

However, if there is any variation, even if the cloud costs the same, the pay-per-use benefit kicks in. Specifically, when U = 1 and A < P, we know that A × U < P × 1 and therefore A × U × cr × T < P × cr × T. In other words, the cloud is the better option.

Under all of these circumstances, where either the cloud costs less or the same, ...