To cloud or not to cloud - that's not really the question!

Earlier this year, I wrote about the cloud diaspora, the idea that companies are moving their IT out of the cloud to save money. As I said then, the story is more complex than it first seems.

Some people might be forgiven for being surprised that this is an issue at all. For ten years, we’ve heard that cloud is always cheaper, usually more reliable, and even better for the planet than on-premises data centers.

There is a counter-argument, that cloud isn’t necessarily the best. And some people put it very simply. Last year, Andreessen Analysts suggested that companies that move their IT loads out of the cloud, could halve the total cost of ownership (TCO) of their IT.

As with the cloud-boosting story, the reality is not that simple. If the question is “to cloud or not to cloud”, there is no easy answer.

Firstly, as I said last time, companies have different IT needs at different times in their lives. You’re best starting with the cloud because it’s easy and lets you focus on your real business. You’re dumb to stay there when you’ve grown large, because it’s costly and hard to control.

But what is this saying? On your company’s tenth birthday, is the CTO going to announce: “OK, everyone. Time to bring the IT in-house!”?

It’s never going to be that simple. For one thing, your company will have a diverse set of applications, some of which are still better in the cloud. That’s the real explanation of the “hybrid IT” buzzword.

For another thing, after ten years in the cloud, your company simply doesn’t have the skills to spin up an on-premises data center. Not without a lot of resources, and a lot of heartache scoping the project.

That’s why those workloads aren’t going on-premises, except for the very largest end-user companies that can afford to build their own site. Most workloads leaving the cloud are now going to colocation. But it’s a whole new generation of colocation.

Applications that were in the cloud, will almost certainly still rely on access to other resources that are in the cloud, whether those are more in-house apps, or shared cloud resources.

And realistically, they also need access to the cloud for bursting capacity in times of peak demand, like at Christmas shopping times.

So any colocation service that wants to welcome the cloud diaspora is going to need on-ramps to all the major cloud-providers. As a minimum, that means network connections inside the data centre which go straight to AWS, Azure and Google Cloud, like those offered by Megaport for example.

Old school colos might have balked at letting in the competition, but that’s the reality now.

They also need to include resiliency. Aside from a few glitches, the cloud is reliable. When a service like AWS has a major problem, the very furore shows how much we rely on these services.

“You can't just have great resiliency,” Greg Moss of advisory firm Upstack told me. “That means big buckets And some sort of bare metal or custom managed component, like Equinix Metal for instance.”

Bare metal is key to weaning your IT off the cloud, because it gives you dedicated capacity, that you can optimise the way you want it - but it delivers it on someone else’s infrastructure. This lets you meet privacy rules, evade cloud exit fees, and still avoid having to become a hardware expert.

At root, there’s a big market for this new colocation. It has to offer elements of cloud, but without the drawbacks and lock-in of the one-stop shop cloud vendors.