ASK IDEAS: How Do You Compare the Prices of Clouds with Different Pricing Models?

Cloud computing is not limited to per hour pricing models. The number of vendors that have introduced cloud products that charge per month is growing, and the seminal Amazon AWS has offered per year "reserved pricing" for some time now. Comparing different pricing models is easy when there are two different vendors both using per hour, or both using per month, but when the units are mixed (per month vs. per hour) the comparison becomes cloudy, pun intended. Equating different vendors is not difficult if built on the proper mathematical precepts. (Namely, "what do I really want to compare?") However, the fact that vendors sometimes use dubious comparisons to make products appear cheaper has led to some confusion on this subject.

How not to compare vendor pricing 

There is one example of how not to compare vendors that is as pervasive in marketing material as it is wrong. If Vendor A charges for the cloud on a per hour basis and Vendor B charges on a per month basis the correct procedure is to transform one unit into the other (e.g. per month becomes per hour). However, the following equation is not necessarily correct:

 

If you failed to catch the mistake you are not alone. Vendors use similar false logic in marketing material to transform monthly pricing models into an outright no-brainer.

The problem with the above formula is that it does not take usage, i.e. duty cycles, into account. If an end-user utilizes the above per month cloud for 360 hours, instead of the 720 hours in a 30 day month, the effective price is twice the value from above. By similar logic, the per hour cloud allows the end-user to shut the instance down for 360 hours, whereas the end-user would lose the cost of the idle time in the upfront monthly price of Vendor B.   In this current example the "effective price" of the per month cloud (in terms of per hour pricing) would be:

 

  

True comparison

The above effective price is straightforward, but it is also the simplest scenario. This blog cannot exhaust all possibilities that end-users will encounter, but there are a few other complexities that consumers need to be aware of. First, although it is possible to compare three or more pricing models simultaneously it is easier to pair off vendors, compare, and repeat until there is a clear winner - like March Madness brackets. Many vendors have mixed pricing units for a single product. A common case of a mixed unit is per hour pricing with an additional upfront yearly or monthly reserved fee. In order to break the pieces into a common per hour or per Month unit the end-user needs to use a modified form of the equation above. Here are two examples:

 

The duty cycle variable in the above equations expresses the usage as a percent. For example, if a cloud is deployed for 360 hours of a 30 day, 720 hour, month then the duty cycle would equal 50%. It is always the ratio of the smaller unit, per hour here, to the converted larger unit, per month expressed in hours in this example. 

Additionally, a vendor who offers per hour pricing is taking a risk that the end-user will leave the service every hour. In a competitive and fair market the per hour vendor must command a premium in order to compensate for these events. Because of the premium, even after all the math is used correctly and the comparison is logically sound, the vendor with the per hour pricing will often be more expensive than the per month vendor.  In those cases, the math can take the end-user to the cusp of a true comparison, but the final decision on whether a per hour premium is worth the extra price is a subjective matter, requiring users to perform an analysis of their own workload requirements.