Cloud repatriation: how do you choose public or private cloud?

Meanwhile, Proximus NXT has also handled a number of cases of companies that are scaling back their dependence on the public cloud. “Removing everything is never the task here,” says Filip. “Companies today usually also use many cloud-native applications, for which there simply is no on-prem alternative.” In practice a hybrid environment is usually still the outcome.

“What’s more, the hyperscalers too are responding to that situation. Microsoft, for example, offers a solution in which you combine public cloud – Azure – with private cloud and manage everything via Microsoft in a single control plane. That makes control of a hybrid environment considerably easier.” Performance can also be a reason to partially leave the public cloud. Solutions that require a lot of computing power or local integration and low latency find that more easily in a private environment.

Companies like Dell and HPE – also important suppliers for the hyperscalers – are responding to this. “They do that among other things in the framework of AI, where powerful machines are needed to train LLMs locally with local data. But meanwhile they offer those on-prem machines too via an OpEx model.” And a company like Netflix, that is by definition cloud-native, also uses a hybrid environment for more certainty in terms of capacity.

However, you don’t just leave the public cloud. “Thorough preparation is absolutely necessary,” says Filip. That starts with an assessment of the landscape. “What runs where, and how complex is it all? We did that exercise for, among others, an importer in the car industry. In such a complex, hybrid environment you don’t want replacement of a single cloud application to paralyze the entire organization. And again: is there, in fact, an alternative for every cloud application that you can run on-prem? You must also study that well in advance.”

Closely connected with this is the analysis of what happens with the data. “A frequently asked question is, for example, whether it’s worthwhile to remove data from the public cloud but have them processed there. High costs may possibly arise because suddenly much more data traffic is needed.” It comes down to carefully studying what the impact of every possible procedure is.

“That way you naturally arrive at the motives behind the exercise. That’s what the decision ultimately depends on: are you looking for lower costs, more flexibility, and so forth.” That analysis should show whether repatriation from the public cloud is, in fact, the right choice.

In fact every business should do that exercise, according to Filip, even if cloud repatriation is not on the agenda today. “What if the price of your cloud contract suddenly goes up by 20% tomorrow? Can you then easily leave the public cloud?” Because although you may save by doing so in the long term, the switch itself naturally also entails costs.

“What if the regulations change and you can no longer use the public cloud for specific tasks? It’s a good idea to think about it and prepare a plan. That way you avoid having to make important decisions at the last minute.”

One of the considerations that a company must make in the context of cloud repatriation is assessing whether it has the right skills in-house, for example to run its own datacenter. “Anyone who has gone all-out for the public cloud for a long time may no longer have the specific expertise for on-prem operations and maintenance. But that too can be solved, for example by choosing a managed service for the installation that runs on-prem.”

And what happens to flexibility? Does a private cloud or your own datacenter offer the same flexibility as the public cloud? “You should definitely include that in your exercise too,” says Filip. “You could opt here for the approach we know from the early days of the public cloud. You run your fixed workloads on-prem, while you can turn to the public cloud for so-called bursting, to accommodate peak times.”

In any case, you shouldn’t take the step from the public cloud to on-prem without due consideration. “Anyone who undertakes the exercise should map out the risks and develop a plan to keep those risks under control. At the beginning of a project in the public cloud a company should also anticipate an exit strategy: how quickly can we remove this workload and the associated data again and what will that cost us?” Legislators are lending a hand with this. New regulations are on the way that will require hyperscalers to make data extraction easier.

The actual switch to a private cloud or on-prem takes place preferably with as little downtime as possible. “That’s certainly an issue,” says Filip, “but nothing that a phased migration can’t accommodate.” And, of course, downtime after migration can also cause more problems. “True, but that risk also exists with the hyperscalers.”

The most important thing is to have a good overview of the total picture. “You have to take a look at the entire hybrid environment,” Filip concludes, “so that there are no loose ends after the migration. You don’t want some of the virtual machines that you wanted to stop using to keep running somewhere in the public cloud.