Home > Workload Solutions > Other Workload Solutions > Whitepapers > Choosing Between On-Premises vs the Public Cloud for Workload Deployment > A Public Cloud computing analogy
Consider the following analogy to illustrate the motivation to move IT workloads from an on-premises location to a Public Cloud provider:
An individual chooses to get rid of their personal vehicle and decides to exclusively use Uber for transportation.
Owning a car for personal transportation entails the following costs:
Meanwhile, if Uber is used instead of a personal car for transportation, it provides the following benefits:
To put this analogy into context, consider that the IT workload = distance traveled.
At the end of a month, in both scenarios, a total cost has accumulated. In the case of vehicle ownership, it will be the total of all overhead and operational costs, but in the case of Uber, it is only the total cost of the Uber rides taken during the month.
If the distance traveled during the month is low, then it is more likely the Uber alternative is cheaper than owning a car, but if the distance traveled is high, it would have made more sense to have owned a vehicle instead of using Uber. In practice, it is not possible to pick and choose which one of the options would be used in any given month as this analogy would require committing to one or the other alternative.
Also, consider that once money is spent on Uber, that money is gone, but money spent on a car provides some amount of residual value. Even though a vehicle is not a great investment it may still provide for some return on investment.
The conclusion is that depending upon an individual’s circumstances and the Distance Traveled, choosing Uber as the sole means of transportation may or may not make sense.
Similarly, depending upon the IT workload in question, it may or may not make sense to run it in the Public Cloud depending on its nature. Clearly, there are some workloads which make sense to run in the Public Cloud. For example, building a proof of concept without committing to purchasing a large infrastructure is an excellent candidate to run in the Public Cloud. Another example is applications (or portions of applications) that need a distributed web presence, since the Public Cloud provides turnkey methods of providing public Internet access to applications.
However, for resource-intensive workloads (consider that this represents a greater distance traveled) such as ML, AI, Data Analytics, Edge Computing, and High-Performance Compute Clusters – these most likely do not make sense to run in the Public Cloud because of the costs charged for the services to support these workloads. Using the analogy, this is like saying the distance traveled is too far for Uber to be an economical alternative to owning a car.