Characterizing Relative Costs and Workload Variability
To decide which approach to use, we need to characterize a few main parameters. The first is the unit cost of an on-demand, pay-per-use cloud resource relative to the unit cost of a dedicated resource, which may be owned and incur a cost for depreciation, may be leased and incur a monthly fee, or may be financed, and represent an ongoing principal plus interest fee to the bank. We will define a “utility premium,” U, as the ratio of the cloud unit cost to the do-it-yourself unit cost. In an example where an owned car costs $10/day and a rental costs $50/day, the premium for the rental car is U = 5, since the rental car costs five times as much.
If U = 1, the cloud costs the same as ownership. It is like getting the rental car for $10 per day. And if U < 1, the cloud is cheaper, which certainly can happen as well. The utility premium, U, may vary over time, due both to improvements or one-time charges in your own operations and to price reductions or increases that the cloud provider may implement. However, here we adopt the convenient fiction that U is constant. We assume that the cost of the “ownership” model for a resource for a unit time is cr and therefore that the cost of the utility pricing in the cloud is U × cr. For a given amount of time, T, the total price incurred for an owned resource is cr × T, and the total price for that in the cloud is U × cr × T. These formulas are just a way of formalizing what we’ve already ...
Become an O’Reilly member and get unlimited access to this title plus top books and audiobooks from O’Reilly and nearly 200 top publishers, thousands of courses curated by job role, 150+ live events each month,
and much more.
Read now
Unlock full access