Follow Jason on Twitter: @jasonh
A cloud infrastructure is defined by three qualities. First, it’s infrastructure as its own practice. Second, it’s highly accessible. Finally, it’s industrialized.
Identifying a cloud infrastructure
Cloud is highly accessible infrastructure. And I know the typical comment people make is “Wait, wait, hold on, cloud isn’t just infrastructure.”
I’m not saying it’s only infrastructure. It’s also a delivery model, it’s a business model. It’s all sorts of things. But if you don’t take a certain approach to the infrastructure, you’re not doing a cloud. The infrastructure is a prerequisite to any of those other things that you may think you want to do.
Infrastructure as its own practice
There are two ways of thinking when identifying this kind of infrastructure. One, if you’re familiar with anything in the enterprise IT sector, it’s not any of that. If anything, the closest thing infrastructure looks like is large scale scientific compute. Operationally, here’s the important part, it is infrastructure as its own practice. You will have a group of people doing infrastructure planning, infrastructure lifecycle, and infrastructure management, but they don’t care what’s on the infrastructure. And the people who are running on the infrastructure simply have to operate within the constraints of what’s underneath them.
Now, highly accessible is the critical difference. And when I say accessibility, I mean it like it’s usually used in design. For example, when you decide to make an iPhone application work for somebody who’s blind, you’re dealing with accessibility.
An accessible infrastructure is an infrastructure that’s easy to consume. It is what clouds are. They’re easy to just go up and use. And you get it.
Now, if you decide to make something highly accessible, it has consequences on every other non-functional requirement of these systems. For example, security. Normally one does security by limiting access. If an infrastructure is accessible, how do you do security? It requires something else. Likewise for maintainability. If you have maintenance windows, but the system has to always be accessible, how do you do maintenance?
Same for capacity planning. If people normally put a request in, and then you calculate utilization, and you do it in a predictable way, and then you buy capacity, and then four months later they onboard on the capacity, that process is only possible because you’re limiting access to the infrastructure. If you allow complete access to the infrastructure, how you look at utilization, how predictable it is, how you do capacity planning, it all changes. So accessibility, just that idea to make something highly accessible and easy to consume changes all these non-functional requirements.
And when we say industrialized, we mean the introduction of supply-chain thinking where there are continuous improvements in technologies, outputs, and unit economics, but all these things occur within a known operational model. There’s not some process where somebody is looking at a spreadsheet six months after the fact and saying “Hmmm maybe it makes sense to do this.” That you actually sit down and say that with the latest and greatest technology, we can follow the technology curve of each of these components, and outputs will go up and unit economics always get better.
The industrialized definition here is no different from the industrialized definition we use in things like radio and mobile networks. Mobile network data capacities have increased exponentially, unit economics have gone down, and our customers have kept their capex spend flat.
So the way that mobile networks have been rolling out for the last decade conforms to the hyperscale definition and looks industrialized. The way that public cloud providers do this conforms to the hyperscale definition and looks industrialized.
The whole rest of the infrastructure out in the world does not conform to this definition. It does not look industrialized.
Taking the next step with Intel® Xeon® Scalable Processor
The launch of Intel® Xeon® Scalable processor, makes it even easier to have both a highly accessible and industrialized infrastructure. It is the most advanced compute core, designed to provide integrated performance across the data center, from on-premise to hybrid to public cloud applications. And it provides “no drag” data protection for security without compromise by lowering the overhead for processing encryption algorithms. So, if you decide to industrialize your infrastructure while making it more accessible, the Intel Xeon Scalable processor will make your efforts even more successful.
For more info about how Ericsson solutions use the new Intel Xeon Scalable processor, click here.