Cloud computing is a group of services. But behind them, there is the supporting infrastructure. Which deployment model should we go for?

Private cloud: the infrastructure is hosted and run from the service consumer; it is a different way to run an on-premise dat center.

Public cloud: the infrastructure is hosted and run from a cloud provider, that offers its services to multiple organizations on a pay-as-you-go or a subscription basis.

Hybrid cloud: combines the two previous models.

Deployment model	Pros	Cons
Private cloud	– Customizable – Consistent performances – Enhanced security / compliance	– Upfront and maintenance costs – HW limitations
Public cloud	– Pay-as-you-go or subscription – Virtually unlimited resources – Delegated maintenance	– Limited customization – Performances may vary – Hidden costs (lock-in, security, …)
Hybrid cloud	– Highly flexible – Cost efficiency – Reduced hidden costs	– Depends on the implementation

Private cloud

Private cloud puts its focus on security, compliance, customization. On one hand, this means that the company has to invest in upfront costs for the data center and the hardware. As well, the hardware scalability is granted as far as there is a good planning and the infrastructure is updated. But, on the other hand, the environment is fully customizable, performances are under control, and not depending on other actors. Security and compliance are fully under control.
Sometimes, dedicated specialists that are needed to run the infrastructure, are considered as a cost. On the other hand, in case of issue, there is quick availability for experts that can work on providing solutions. Solutions that, if delegated to a third party, are going to come in accordance with their business plan, not the business plan of the user. And we know that, in IT, problems are simply a matter of time, we know they are going to happen.

Public cloud

In the public cloud, upcost for building the infrastructure, for owning the data center, maintaining it, it is all delegated to the service provider. That, in exchange, can offer pay-as-you-go or subscription tariffs. The idea is that you are going to pay only for what you use, when you use it. In theory, this model is supposed not only to avoid to spend for acquiring resources, but also for paying for what is effectively used, optimizing the budget. Resources are virtually unlimited and considered to be available at any time, without need for scheduling, reservation and so on.
All this, is not coming for free. Customization is limited to what is allowed by the service provider. Performances may vary according to several factors, from the public internet to the amount of effectively shared resources, for which there is a competition between all the consumers served by the provider. But there are other hidden costs. Security, is a factor to be considered. A data breach can impact multiple customers, and there is no guarantee will be detected. Also, it is true that the customer does not need to allocate budget for its own data center. But this means he does not have the full control on its data, processes, workflows. The customer, once has put his information in a container he does not own, is subject to the service provider policies. That may involve changes in conditions or service levels, change in costs, forced downtime. Therefore, some of the supposed costs coming with a private cloud, or a private data center, may become hidden costs in a public model. Other factors should also be considered. Not all applications are optimized to work well in the cloud. This means, that one of the main factors that allows to optimize costs in the cloud, may not be applicable, resulting in having to deal with higher overall costs. And again, we have not discussed about compliance to regulation. Laws can change from country to country, and some countries may put constraints related to data sovereignty, complicating the management of data.

Hybrid cloud

Partially due to these reasons, partially to simplify the migration, there in a third model, the hybrid cloud. In some cases it is seen just as a transitional model. Utilized while migrating from one to another. But, if we consider previous points, we see there is a good reason for it to exist on its own. The idea is to integrate the other models, allowing to create an environment that combines the advantages of both. A highly flexible environment, supposed to be cost effective and to reduce hidden costs, for example keeping strategic flows internally, and reducing the possibility to fall in the vendor lock-in trap.
Cons of this model can vary a lot, according to how much is kept in the private infrastructure, and how much is deployed in the public cloud.

Conclusion

As always, in engineering the reply to “which model is better” depends on several factors. Often, not always immediately clear. Each model has its advantages. But considering the overal situation, with global instability and sovereignist pressures, we may see changes in organizations plans.