Traditional private clouds are on their way out. I’m talking about the open standards-based stuff that was huge a couple of years, but fell behind in features or functions compared to the public cloud providers. Other factors included complexity of installation and configuration, as well as the fact that you still owned hardware and needed data center space.
I still consider the concept of a “private cloud” a useful architectural pattern, in some cases. Some enterprises have special needs where traditional private clouds still fit—few and far between, but they do exist.
If traditional private clouds die off, what replaces them? Here’s my take.
First: on-premises public cloud extensions. I’ve also called these public cloud peripherals. They are typically proprietary, on-premises private clouds that are purpose-built to work with public clouds. Two in the industry now include AWS’s Outposts and Microsoft’s Azure Stack.
These are designed to provide an on-premises platform for workloads that cannot (or more likely don’t want to) run on public clouds, but want to be public cloud-platform ready. They use many of the same native APIs and should provide A-to-A portability with their public cloud counterparts.
They are a good way for public cloud providers to gain access to the market that’s not yet ready for public cloud but wants to begin the migration on-premises. Thus, this approach is partially marketing focused.
Second: middleware-connected traditional systems. If we’re not moving on-premises workloads to private clouds, we need to make the traditional systems, even mainframes, function as private cloud analogs.
We do this by connecting the on-premises systems to public clouds using middleware. This allows the traditional systems to share processes, data, and so on, and work closely with public cloud providers, even partitioning applications and data to run on both places for a single system.
Of course, there are other traditional private cloud alternatives that I’ve not listed here, but these two methods are the most popular.
I suspect that moving more workloads to the public clouds means fewer on-premises systems. That will also remove the need for new on-premises platforms, including both solutions I’ve listed above. In that case, they are stopgaps, and that’s not a bad thing.