My Amazon author page!!!!

I think about cloud topics all day long sometimes. The one that caught my attention yesterday is cloud pricing.


The why of what I think about would probably involve lengthy therapy sessions but for this one its because the model is interesting and fits with a project I was working on.

There are three accepted cloud pricing models today, one-time, utility and subscription. Each has factors that weigh in its favor and downsides as well so let’s take a look at each side.

One-Time pricing represents an agreement for a time period to purchase a set of compute and storage resources. These will be used for the agreed period and then completely removed. This could be an hour, this could be three weeks it is simply a contractual vehicle that supports using assets once and then releasing them. The upside is you can buy a lot of compute power for less than buying a Quantum or Super Computer. If you only need these type of power once or twice a year it is significantly cheaper than investing in a large compute or HPC system. The downside of this is if you are doing something with large data sets your one-time period has to be longer an therefore more expensive than if you are doing this on small data that builds up quickly.

Utility pricing represents the promise of the Internet it has a number of different names pay-by-the-drink, pay-as- you-go etc. In college we used to pay by the drink until about the 4th one then in the morning we paid for the drinks again. That is the reality of utility pricing. It can be great if you solution is a cloud application and the value proposition of cloud fits right into what you are building and delivering. The risk? Applications gone wild. Bad apps when moved to infinite resources can become very expensive. In the days of pre cloud computing we were limited by the hardware available. With the cloud that application can expand forever.

Subscription pricing has gotten a lot of bad press and frankly for some organizations is well not deserved. In fact if you have known compute requires this may be the best option for your organization. it provides you with access to a dedicated environment dedicated to your organization. The down side is if you over power your subscription you end up with a higher cost than needed and nothing is worse than wasted cloud processing.

In this blog I would like to propose a new model. That being the combined Utility and Subscription model (which in the end is where most companies end up). The difference here being that you negotiate a favorable storage subscription and a baseline compute subscription that covers you daily compute and storage needs. You then negotiate based on this new volume an ongoing Utility pricing model.

In the end Caveat Emptor. What you buy should be based on a careful evaluation of what you really need.


Scott Andersen

IASA Fellow