Are the days of renting a room in a data center, buying servers and having engineers nearby to solve potential problems gone? While building an on-premises infrastructure used to take weeks (or months), it now takes hours. With prices ever lower, the question is simple: does on-premises infrastructure have a future?

A skill management problem

Cloud computing has established standards for deploying software, starting with EC2, S3 or Kubernetes. Not only are these technologies very well understood and mastered by engineers (after all, they have learned these technologies during their professional experience or during personal projects) but their use in cloud computing platforms is facilitated with tools mature and well supported.

Conversely, when you deploy your own infrastructure, it becomes necessary to develop specific tools to manage your servers and the deployment of the software. It then becomes a full-time job that requires hiring engineers. In addition, this specific tool will be less mature than their equivalent available for the cloud and therefore slows down the company in the development of its infrastructure and its applications.

So the choice is simple: by using the cloud to build your infrastructure, you can have access to a mature platform and working candidates from day one. Choosing on-premise means a long period of hardware installation but also the hiring of engineers who will define the software deployment methods and the implementation of the tools.

Ever-increasing performance, ever-lower costs

With adoption growing by 30% quarter on quarter, cloud computing players now have such a large user base that they can optimize their infrastructure themselves and provide their customers with ever more efficient services at ever lower costs.

The perfect example is Amazon Web Services’ development of the Graviton, a processor that delivers superior performance to Intel processors at a lower cost. This technology is of course limited to Amazon Web Services customers. By developing its own hardware, AWS is able to remove all non-value-added costs and focus on developing a successful product for their customers. These optimizations apply to all infrastructure deployed in the cloud. This processor is not only used to deploy instances of your software but also for hosting your database or orchestrating your infrastructure. These optimizations are far from negligible with 25% to 40% higher performance and a 5% to 10% cost reduction. Choosing to stay on-premise also means refusing these potential optimizations.

Security: a scarecrow argument

An argument often raised against cloud computing is the potential security of data. Cloud computing providers are considered insecure infrastructures and using an on-premise infrastructure would ensure data security. A simple google query shows that security is top of mind when it comes to cloud computing.

However, thinking that on-premises is more secure than the cloud is like thinking that driving in a car is safer than traveling by plane because we are in control. However, the vast majority of security issues in the cloud come from usage. Configuration issues are legion (for example, S3 buckets being incorrectly configured as public or misconfigurations of routers and load balancers). But these are easy problems to fix. Many services are available to scan an infrastructure deployed in the cloud and find any potential security problems (which is much more complicated to do for an on-premise infrastructure).

Many companies with security constraints use the cloud today. Capital One (a bank established in the United States) has migrated its infrastructure to the cloud. Even the US government today uses cloud computing with services dedicated to them (and this trend is not going to stop since Pentagon announced a few months ago a new contract to develop their infrastructure in the cloud).

The security argument is only a pretext, a psychological barrier that some decision makers use to not change the status quo but which exposes their company to be less innovative in the years to come.

What future for on-premises?

There is little doubt: the on-premise will become a niche market in the next ten years and new cloud computing providers will emerge. However, there are still some minor barriers that can force companies to stay on-premise.

First, the inertia of old companies that have a well-established on-premises infrastructure. These companies already have premises (and rental contracts in data center) and servers and are unwilling to change the way they manage their infrastructure, citing that their current infrastructure is so well-honed that switching to the cloud will cost them far too much. However, by keeping this old approach, they will end up paying much more in the cost of developing and maintaining their infrastructure (engineers dedicated to managing the infrastructure and its tools) and the cost and performance gap between cloud and on-premises will continue to grow.

Second, a guarantee of being able to switch service providers. Many cloud solutions are vendor-specific (for example, using an Amazon Aurora database on AWS does not guarantee failover to Google Cloud Services or Azure). This is a real problem that can be avoided easily: by using generic and interoperable solutions, it is easy to be able to change service providers. Instead of using ECS ​​to deploy Docker containers, it makes more sense to use Kubernetes; instead of using Amazon Aurora for its database, it makes more sense to use MySQL or PostgreSQL. To avoid this problem, many companies now facilitate deployment and migration between service providers.

Finally, a problem of sovereignty. In other words, where is your data physically located. This is a critical point for Europe where no cloud computing provider can claim a quality of service equal to Amazon Web Services or Azure. While this does not affect small and medium-sized enterprises, it is however a critical point for large enterprises or public institutes. Unfortunately, European initiatives that tried to copy their American counterpart ended in failure. (who remembers Quaero?). However, the technical skills are there, developing a European cloud worthy of the name is only a question of political and financial will.