VMware Lightboards - Intro to Kubernetes, 16 Jul 2018

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Mapping Kubernetes to your Infrastructure

Description

In this lightboard video, Steve Tegeler guides you on how to map Kubernetes to your Infrastructure, using Software Defined Data Center examples from VMware.

A

Hi, my name is Steve Pegler I'm, a senior director systems, engineering and VMware's cloud native apps business unit, and this lightboard session is about mapping kubernetes to your infrastructure. As we all know, kubernetes provides some great primitives to work with, and an architecture to define applications, that's fairly straightforward and easy to work with lots of ecosystem.

A

Examples are out there, and so what I want to talk about in this session is how do I take those primitives that exist and the definition of what I want to do and how do I actually do that and how do I do that on a given infrastructure? So, if I think about the active she's up here and she's, let's just say, she's going to create an elk sack and part of that elk stack is she's. Gonna have to define all the different things that need to intercommunicate.

A

Maybe some require persistent storage, maybe there's some security somewhere right, so she's, basically gonna architect. This and she's gonna architect. This leveraging the kubernetes api in the kubernetes primitives. So inside kubernetes I've got these various primitives I've got availability zone, defining security policy load, balancer I've got persistent volumes here, and then you know metrics. So after I deploy my app, how am I gonna understand how its operating and working, and so the idea here is that you know what I look at say, something like persistent volumes.

A

The definition exists in kubernetes, but what I actually have to do is I have to store that data somewhere right. It's got to end up on a spinning disk, probably somewhere or an SSD somewhere in my in in my data center, in a public cloud.

A

So what I'm gonna have to do as a part of this is I'm gonna have to map the the ability to create persistent volumes, say, maybe via storage classes, down to an actual storage infrastructure solution, and so that's exactly where our traditional infrastructure operators step in so normally or what has been method here, is that the infrastructure operators are responsible for infrastructure. What we want to see these infrastructure operators do is we want to start having them elevate themselves?

A

We want to have them move out of just infrastructure and we want them to become prop form. Reliability. Engineers basically start leveraging some software principles and software methodologies, but apply that to my traditional platforms that I provide internally to other teams. So a platform reliability engineer now, in this specific case, let's just be very specific. This is really you know a kubernetes cluster operator and they're gonna be responsible for this mapping.

A

Of course, persistent volumes availability zones- you know if I'm, if I'm leveraging on-prem infrastructure, I'm gonna need to map the availability construct here that exists in kubernetes down to a group of servers in my data center in a specific rack in a specific row, I'm gonna need to do that same for security policy. I'm gonna need to take the security policy and I'm gonna need to map it into some sort of technology that can enforce the policy that I define in kubernetes load, balancing same thing and then metrics right.

A

What's in, what's gonna be an easy way where maybe, as I, deploy this ELQ stack into kubernetes by you know, just by defining the fact that I need some metrics, it will automatically happen for me, okay, so that is that's kind of the idea of what we want to see here. So you know being from VMware. What I'd like to talk about is the way we map to VMware constructs and mainly the software-defined data center. So in this case the compute is obviously vSphere.

A

So when I think about availability zones, those will get translated to vSphere clusters. When I look at security policy, those will get translated into the NSX distributed firewall, so we'll be able to provide that micro segmentation down to the pod level within kubernetes load. Balancing same thing, it's nsx we're going to leverage the NSX load, balancer type load, balancer, persistent volumes.

A

Here we can leverage traditional data stores or we can leverage just visa and we can get some advanced functionality like you know some D dupe capabilities and encryption so forth, and then finally, monitoring the technology here from VMware is wavefront. And so what we'll be able to do is we'll be able to get a little piece of code inject it into the deployment file here so that when they deploy their elke stack, it automatically starts collecting information and metrics about about. What's going on in their app now I've just defined.

A

You know all of the VMware constructs here now, certainly in each one of these there's a value prop with just vSphere leveraging it with kubernetes and just NSX and just wavefront, and just you know the storage solution here, but the reality is I'm. Gonna need to map and I'm gonna need to configure each one of those up into kubernetes. Wouldn't it be nice if, because I knew this was going to be the software-defined data center I could have a very prescriptive approach.

A

In terms of how these things are mapped- and that is exactly where PKS or pivotal container service comes into play, so PKS can leverage all of the software-defined data center here and we can automatically create this connectivity so as we deploy PKS we'll just feed it information about where every Center is information about where the nsx controllers are and once that initial definition is done as I do deployments of kubernetes.

A

These linkages will automatically be done for me, huge power, so there's there's value in the individual technology, but PKS brings it all together and creates a consistent and repeatable method for deploying a kubernetes cluster. Now I've only shown the VMware software-defined data center PKS can also deploy to various public clouds. At the time of this, recording PK bks can deploy to GCP, but look for other public clouds in the very near future.

A

So now we've kind of satisfy the need of both the platform, reliability, engineer and the application developer from a platform reliability engineer if you're leveraging PKS, you have a very consistent repeatable way to deploy kubernetes, but you also has an easy way to map it. You have the ability to do upgrades and patching in a very easy and supportable way so boy, these platform, reliability, engineers they're pretty happy from an app dev standpoint.

A

What do you have well, you've got your native kubernetes api PKS simply deploys native upstream kubernetes, and so in that case, she's gonna be happy, because not only does she have a native kubernetes api vic, she can leverage all the ecosystem of information around, but she knows be. You know, because we've got this easy mapping down here to a known infrastructure. We're gonna have a pretty reliable API, that's going to be up and running so hopefully this gives you some good understanding of how how all this works and how it all fits together.

A

Thanks for watching.