Read more: |
Matti Grönroos
Cheap as Cloud
The "term" cloud is defined in quite a loose way. Fortunately, things have changed since the peak of the hype, when everything related to networking was cloud.
The cloud vendors have succeeded in what many have tried: They run wholesales for the retail prices. That is why every customer shall be careful.
Correct.
Especially, when there is a need to invest in a new data center, or make a major renovation to an existing one, the IaaS is a pretty evident choice. However, it is not self-evident if the business case is good or not, and the calculation is not easy. Damn low unit price times a damn large number of used units can produce a bill of a hell of a lot of euros.
In its simplest form, the math is a piece of cake:
From the pricing perspective, the IaaS is close to the IBM Mainframes a generation or two back: The glass house was selling CPU seconds, I/O requests, disk space in cylinders and tracks (19069 bytes on a 3350 disk, some other number on another model), magnetic tape minutes, line printer lines and pages, read and punched cards, etc. A complex price list might have even four pages on those days. I don't think four pages are enough for Azure's price list in 2024. What is like in the mainframe dinosaurs, is the customers' need to optimize the consumption profile to match with the pricing principles. Failure to do so could result in a surprise bill.
Before widespread server virtualization, it was estimated that 80-90 percent of the installed capacity in a typical data center was without load (but not without electricity consumption). Virtualization turning mainstream has changed the situation, but the application base may be very problematic in terms of cloud pricing. In the application where the WWW server is on 24/7 waiting for a few transactions per day, the price per transaction would be something that the CFO may not necessarily love.
In practice the application cloud transition is not necessarily a simple exercise but a major project. The applications may need an upgrade, and the back end potentially needs a full rethinking. In addition, restrictions related to data protection may be encountered, which makes the transition difficult. It is not uncommon that customers need to have a small data center of their own for those services which cannot be transferred to a cloud.
The trend is towards PaaS and SaaS, Platform as a Service and Software as a Service.
The PaaS model has fascinating advantages because, among other things, it frees up a large part license management seen as cumbersome. In addition, for a relatively small amount of money, the customers can get access to software whose licenses would be too expensive for them to acquire. The pricing of the PaaS model is even more complicated than that of IaaS, thus making it necessary to get familiar with it and its pitfalls.
Another thing to note about PaaS is that it implies a commitment to the cloud provider's rather strict program versioning. The working model to apply version updates to the platform software every few years only, does not work. This may be a problem to software vendors: New versions are needed even if the lifecycle of the software is approaching its end.