Wrong expectations
Advertising suggests an image that the cloud is definitely cheaper. Unfortunately, this is too generalized and the application case should be exactly determined and considered with which one would like to go into the cloud.
If you enter Cloud + cheap in Google as a search term, you get a lot of supposedly cheap offers and unfortunately the idea of the free cloud has also got stuck in the minds which, as we all know, is paid for the access and use of our data.
Different service models
If you look at the different types of cloud - IaaS, PaaS and SaaS - you can quickly see that in a pure infrastructure cloud (IaaS), a high proportion of responsibility remains with the customer. For example, support for the operating system, middelware, user support, etc.
More relief is offered by a platform service which, in addition to the usually fast provision, frees the customer from the support of the operating system and the middleware. A not insignificant saving. Databases and software development environments are thus provided in their entirety and billing is often based only on transactions or duration of use.
Last but not least SaaS services, which we are already used to from our smartphones, the apps are maintained by a team and I only consume them on a per use model. Example of an ERP system where costs are only incurred for issued invoices.
Incorrect workloads
The prime example, unfortunately seen too often, is to move a classic legacy workload to an IaaS cloud and then run it there 7x24 hours. This often happens for example with Exchange, an application that has to run 24 hours per design. What should be the savings through cloud computing? Cloud is based on elasticity, workloads that have different needs over time and can also be stopped.
Cloud is by design evergreen, there is no need for classic methods to achieve fail-safety, it requires up-to-date software to benefit from the mechanisms of the cloud.
Does Cloud really have to be cheaper?
A question I often ask, do unit costs really have to be cheaper than on-premise? Is the rental car cheaper than the car at home on holiday in distant Australia? Certainly not, but it allows me to be mobile during my holidays. Enablement is an important factor in the cloud. It enables SMBs in particular (and that is 98% of our companies) to obtain services they could not afford before, for cost reasons such as Skype. Which SME could afford a Skype infrastructure a few years ago, today this is no problem, as a SaaS service from the cloud for a few euros per month. Also, if the skills are missing, does the company have the team to develop and operate the Service? When it comes to IoT or AI, specialists have to be employed first, therefore it is easier to fall back on finished stacks running at a provider. Besides the'scale economy' that many expect from a provider due to its size (keyword hyperscaler), one must also speak of the'skill economy', you receive with the cloud a troup of specialists who work for the customer and which someone could never finance in his own company.
Lack of cost control
Another factor that can lead to the valley of tears is the lack of control over the costs involved. An account with a provider is created and then resorces are booked happily. Resources that are no longer needed are not switched off as usual and the costs continue to run.
Of course, it must be noted here that cloud and on-premise comparisons are often not made like-for-like, but that internal costs are often dismissed or deliberately negated. To save departments or jobs. In addition, so-called 'anyway-costs' costs are often not included in the calculation. An employee who does user support in addition to his job in accounting, because he knows Excel so well, these costs are usually swept under the table.
upshot
When considering placing a workload in the cloud, it should be carefully considered, among other technical and legal aspects, in which service model of the numerous XaaS variants this specific workload is optimally balanced. With legacy software, one should look for possibilities to find an adequate solution as SaaS with an accounting model with which the costs go along with the incomes.
Many shun the costs of reprogramming, but if a modern architecture is chosen, they can be quickly made up for by a much cheaper operation.
An external consultant, who may have already calculated many use cases, helps to get an objective picture of the matter. In the Master of Science in Cloud Computing Engineering course at the FH Burgenland, the FiMeK module (Financial Mathematical Methods and Calculation) provides exactly this full-cost analysis.