So, containers have taken the world by storm over the last four, five years, but I wanted to take a little bit of a look back maybe for those of you who are only just getting into Docker or containers or Kubernetes or perhaps to provide a brief review for some of you who've been living it for a while of why this whole thing came into existence in the first place.
I think to understand why, you should cast yourself back into the world of cloud circa 2012.
The truth is that for most of us, we were managing virtual machines, and they were cluttered things, they had all sorts of different components on them.
They had SSH servers, they had monitoring, they had your application itself.
It was hard to know exactly what was going on, and the truth is that most of the time when we're managing them, we're managing them with some imperative script.
Maybe this was a Bash script, maybe it was something like Chef or Puppet, but in general, it was something that we ran on the VM.
Maybe it worked, maybe it didn't work.
It was difficult to understand if it failed, how to reapply it, and everything else like that.
In this context, there was a move to what became known as immutable infrastructure, and the first way is that immutable infrastructure came into existence was to say, let's actually just capture this whole thing into a VM image.
This way, by using a VM image, everything is immutable, it never changes and always deploys exactly the same way.
I don't worry about how did I make certain changes.
Did I login at the command prompt and type something like apt-get? I know exactly what's going on, I don't do any of this stuff.
But the VM images themselves, while they were great from an architectural perspective, they were really heavy from a developer perspective.
So, into that world came the idea of saying, hey, let's take that VM image actually and let's turn it into a container image, much lighter weight thing.
Instead of being hundreds of megabytes, maybe a good container image is 20 megabytes.
We'll separate out the app from the kernel, and there'll be this distance where I can say, you know what? When I care about my application, I think about my container image.
When I care about the operating system, I think about the kernel.
That decomposition and that separation of concerns really made the artifacts something that was worth, which made it easier for application developers to think about building and deploying the application.
But that wasn't the only reason, just being able to separate out your application from the rest of your operating system.
There was another part of what Docker supplied, which was the cloud repository.
So, in any real application, not only do you have to build and package your application, but you actually have to distribute it around the world.
So, I have a laptop here where I've actually built up my app.
But how do I get it to a data center that's connected to the network? How do I get it to the cloud somehow? How do I get it to multiple clouds all around the world? Sorry, picture of the world here, it's not working out so hot.
How do I do all of that? Well, it turns out that the container repository provided exactly that capability.
So, you're going to have a repository like the Azure Container Registry or many other container registries or the Docker container registry.
The developer can push their container image up, and they can pull that image wherever it needs to be: into their data center, into the cloud, or even around the world.
So, now, we have this immutable infrastructure, so we can get rid of all these scripts, we can have guarantees that your application's going to deploy correctly every single time, but we've also gained a way of distributing that same application all the way around the world.
Finally, the container image itself also became a way of putting multiple applications onto the same machine.
So, if you have a VM here or a physical machine as well, one of the challenges that a lot of people run into is that on average, it's something like 10 percent utilized.
That obviously means you're paying for way more than you're using.
One option is to make better applications.
That's often hard to do, specially with certain languages that may be single threaded or have other constraints, legacy applications that you don't want to re-architect.
If you package up your application as a container, it's actually relatively straightforward to put app one, app two, and app three all on the same machine, drive that utilization up to 50 or 60 percent without doing any sort of re-architecting at all.
Now, it's important to note that this boundary here is a resource boundary that's defined in the kernel.
It'll give you a good way of isolating in terms of things like memory, that's my memory chip drawing, as well as in terms of CPU usage, but it is not, and this is important to remember, it is not a security boundary.
So, it's not going to protect you against malicious code that's in app one from affecting a code that's in app two.
If you're trying to use containers as a security boundary, there's a lot of other additional technology you need to add on, something like a hypervisor or other kinds of isolation technology.
But if what you want to do is be able to drive up your utilization for a bunch of applications that you trust, the container can become a great artifact.
But to really make it something that is useful, you need to use a container orchestrator.
That's where something like Kubernetes comes in.
So, we have a whole bunch of machines.
We can use the Kubernetes API to distribute that container image or a lot of container images onto all of the various machines, but we'll talk about that later.