Lately there’s been some confusion: places like SIDN (Dutch national operator of all internet names that end on .NL) claim that nobody in Europe can deliver their computer needs, and that they therefore must outsource their operations to American cloud providers.

Dutch version of this page here: Cloud Native, Europa, de ‘Bijenkorf’ Megascaler 🇳🇱

Meanwhile, we have large providers of servers and services in the Netherlands and in Europe who claim that they can (and do) provide these kinds of services just fine. What’s going on?

In this piece I want to sketch the cloud situation and specifically Europe’s lamentable position. As I quoted an EU official earlier, “Interdependence is natural, even desirable. Over-dependence, however, is not”. And Europe is very much over-dependent on specific technologies from other continents. Even if these are our allies (for now).

IT services run on software, which runs on hardware. When we zoom in a bit, we see that many software programs do not stand alone, and need further underlying services, such as databases.

A provider of a service (for example SIDN) can deliver such services in multiple ways. One way is to organize your own software, buy or rent hosted servers somewhere, and run your own software and also databases and other underlying requirements on top of that. This is the traditional model, some would say: the historical or even “old fashioned” model.

A new way of developing is called “cloud native”. In this approach, a provider focuses solely on their own software and runs it on top of services from third parties. You don’t run the database, for example, yourself; it’s provided “as a service” by a third party. And possibly many more parties provide underlying services. Your software no longer runs on hardware or servers, but on services.

Cloud native is not well-defined, but an operational definition is “building on Amazon Web Services, Microsoft Azure, or Google Cloud Platform”. This approach comes with a set of unwritten assumptions about what these three parties then should offer.

Part of this is a vast array of services that is automatically available and can be controlled programmatically (“infrastructure as code”). Additionally, cloud native users expect a self-scaling infrastructure that (if need be) recruits more computing power (“Managed Kubernetes”).

The largest three cloud providers (also known as “hyperscalers”) can be seen as “IKEA’s”: they offer everything and it’s almost always immediately available. This was made possible by an enormous investment both in terms of money and innovation. I wrote about this earlier in “Taking the Airbus to the IKEA Cloud”.

The Confusion

When a company or organization looks for a cloud-native provider according to the informal definition above, you’ll find nothing but disappointment in Europe. Our large service providers (Hetzner, Leaseweb, OVH, Ionos, Scaleway etc) offer good servers (computers) and some extra services, but not even a minimally complete “cloud-native” package.

Operators like SIDN can therefore genuinely claim that nobody here can meet their requirements. By the way, it is important to note that they themselves have chosen this (new) ‘cloud native’ way of operating. They did not have to do that.

Our European providers, on the other hand, correctly assert that they can indeed run the required technologies and do so at scale. Even over at our providers, large databases and managed environments are available. And also quite a bit cheaper than “the big cloud providers” (who can be astonishingly expensive).

The confusion revolves around the way in which this is all delivered. The American providers offer this kind of service without any human intervention, and it all happens frictionlessly, in an automated fashion.

Here in Europe, you’ll have to go through an intermediary that sets up the desired services for you, on top of the advanced server (computer) and network infrastructure that we do have available here.

A comparison that might help is: the hyperscalers from the US are like fast food restaurants with very clear menus full of things they can deliver rapidly and reliably. You can even order through an app.

And in Europe, we have catering companies that can make everything for you upon request, and wholesalers (like Leaseweb, Hetzner, Ionos etc) where ingredients can be bought or rented.

The Denial

A problem here is that many European providers have chosen to be a wholesaler of servers and storage, and at the same time, they think they can compete with hyperscalers, and even send angry letters demanding that American providers are excluded from tenders so that they can get a chance.

Partially because of a useful meeting with the Dutch Minister of Economic Affairs and the State Secretary for Digitalization, there has been a lot of contact in recent weeks with cloud users, European server companies, and (hyperscaler) providers. This showed painfully that European companies need to take a closer look at what “cloud-native” actually means.

A significant part of the most essential cloud-native technologies is simply not offered (well) here. And worse still, many providers here have no idea what those essential technologies even are.

To make it concrete, we are talking about:

S3/Google Cloud Storage/Azure Blob Storage
Solid and reliable managed Kubernetes
Infrastructure as Code, Terraform/OpenTofu/CloudFormation
Identity and Access Management (IAM)
Managed databases (“just a database” and to some extent Aurora, Spanner)

Some of our own providers do offer some of these services, but often with significant limitations in geographical distribution, scalability, performance, or reliability. Amazingly, a large cloud initiative here in Europe was not aware what S3 was, even though it is likely the most essential service.

There is therefore a fundamental dichotomy: if you want to work with an IT intermediate, everything can be organized for you in Europe (“catering”). If you want to buy/rent servers and basic services yourself, you’re also well at home here in Europe.

But if you think like a “cloud-native” and want to work that way, you find a European industry in disarray, an industry that claims it can do everything you need, but at the same time doesn’t really know what that means. An industry that can’t deliver automatically and on a large scale your basic requirements, but where there are caterers who (credibly) can make something special for you on request. But that’s not what the cloud-native world is looking for!

By the way, if you start with “cloud-native” without preparation, it often ends in tears. Without expertise, it quickly becomes chaos and extremely expensive. The shared responsibility model of hyperscalers means that they don’t solve all your problems, including certainly not your security challenges. Large cloud providers are far ahead of us, but they are not magic that solves everything for you. They can also have their own security problems.

The way forward

To start with, places like SIDN need to think carefully about whether their choice for cloud-native hasn’t already ruled out a lot of options. It’s about delivering good IT services, and by talking to our existing industry, you can very well roll out those IT services. By specifically choosing cloud-native with the definition “like what Amazon and Azure do” you’re indeed choosing Amazon and Azure.

At the same time, our own industry needs to have a serious look at itself. There is great demand for working in a cloud-native way. It’s not feasible to suddenly also offer all the services of (say) Google Cloud Platform like you can’t suddenly set up an “IKEA” store out of the blue (although some are trying).

But it’s also not acceptable that the players here who want to join in don’t even know how to automatically and at scale deliver the most elementary of cloud-native services.

The “Bijenkorf Megascaler Cloud”

Apart from IKEA, there are smaller but still thriving furniture stores where you can also find excellent items to buy. Institute Clingendael, a think tank, wrote in this context about the Bijenkorf Cloud Megascaler, in contrast to the existing hyperscalers:

“An attractive European proposal requires an ‘All-in-1 package’ – including storage, databases, security, and software development tools – as an alternative to existing American proposals. Only by working together can Dutch and other European cloud providers come up with such a European ‘Bijenkorf Cloud Megascaler’.” – Netherlands and the EU: Go for cloud sovereignty, Clingendael (Dutch).

(The Bijenkorf is a bit like the Dutch Harrods, except somewhat less so. Or compare Nordstrom in the US).

Photo by Renate Dreyer, edited by Guus Hubert

It is well imaginable that European providers do not immediately try to become hyperscalers, but instead start by offering a credible and solid package of essential cloud-native services. This may not be very attractive to the business world that has no or fewer sovereignty concerns and simply prefers to do business with the largest providers.

But those who want to be in control, or sovereign, should find the “Bijenkorf megascaler” a good thing to build on. You don’t always need a whole IKEA. And in fact, the complexity of the hyperscaler clouds is now so daunting, some simplicity (also in billing) may in fact be very welcome.

To form such a megascaler cloud, European companies should work together intimately. Only then can they quickly reach the critical mass that is required. Since our own industry is relatively small compared to the hyperscalers that dominate the market entirely, this does not lead to competition concerns.

There’s also an abundance of software talent in Europe to build services around these kinds of things.

This may be surprising, but it turns out we have a lot of wasted IT talent in Europe.

Governments could play a large role in stimulating such collaborations, not least by credibly stating that they will be doing business with these European megascalers.

Another important role for governments is helping to identify what the essential cloud-native services are, sorted by importance. If clarity exists on this point, this list can also be included in tenders, instead of the current problematic texts that effectively order “Microsoft Azure”.

A good example of an initiative to agree on standards is the Dutch municipality’s Haven standard. This standard lists specific Kubernetes requirements that Dutch cities and towns can build upon.

With more such consultations, these kinds of standards can be expanded, and a healthy market could emerge where many (existing) parties can function as providers.

An additional benefit of such standardized services is that buyers could also get out of an existing provider. Something that is virtually impossible once you’ve committed to a hyperscaler. The Bijenkorf mega-scaler would do better in this regard as well.

More Information

More can be found in two contributions by Instituut Clingendael, as well as my own blogs:

Underlying this piece is a much longer and more technical article titled “The Cloud Gap in Europe” which I am still working on. Anyone interested, please feel free to reach out to me at bert@hubertnet.nl - preview readers are very welcome!