Cloud Native Computing Foundation KubeCon + CloudNativeCon North America 2022, 11 Nov 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Keynote: A Glimpse Into the Future of Cloud Native: Emergent Infrastructure - Frederick Kautz

Description

Don’t miss out! Join us at our upcoming event: KubeCon + CloudNativeCon Europe in Amsterdam, The Netherlands from April 17-21, 2023. Learn more at https://kubecon.io The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy, and all of the other CNCF-hosted projects.

Keynote: A Glimpse Into the Future of Cloud Native: Emergent Infrastructure - Frederick Kautz

Speakers: Frederick Kautz
We are at the edge of a new computing paradigm. Application developers and operators are replacing sysadmins as the primary focus for runtimes and tooling.

New runtime environments are entering the ecosystem, built on the lessons learned from this community, which will increase the total capabilities of developers to build, scale quickly, and deploy applications. These platforms are dropping support for APIs such as POSIX in favor of more straightforward and scalable approaches that focus on the application's needs rather than the kernel.

In this talk, we will explore the long-term future of Cloud Native applications, their implications on infrastructure, and how we can rise together to meet the demand.

A

So, thank you, everyone for being here and thank you for making it an amazing kubecon. So um before we jump into the topic, uh there's something really short: I want to jump into, and first uh this really is about the community everything you see here, like everyone that is working on all these projects. It's at the bottom. We have people that are here so like first round of applause for everyone who is here who is part of the community and for everyone who is not here who is unable to attend.

A

Thank you and, and one other thing as well, is that you'll see a photo of this auction make a plane of this pretty pretty soon, but if you are also in need of like if you're, not if you're feeling burnt out, if you need help like please reach out to somebody, please look after yourself like we don't.

A

This is not a Sprint, it's a marathon, and it's something that we want to make sure everyone is doing things they can not only for the best of the projects, but also for the best of yourselves, so make sure that you are also looking after yourself, drink lots of water. Get lots of sleep, make sure that you're enjoying yourself and help make it a safe space, not only for yourself but for others. So again, thank you.

A

So, let's jump into it, there's this concept that we have called emergence and so emerging system. So it's when you have a uh a system where the individual components that make up the system do not actually have the properties of the system as a whole. They have unique behaviors that arise as part of that system.

A

The first example is: uh is you have this ant, so you have a flood that comes along the Ansel clustered together and now they float and last I check ants, don't tend to float very well and survive, but in this scenario they gather together.

A

You also have all the pheromones that they put out like the individual like how do they go off and gather and act as a whole, like they have entire behaviors they're, not actually encoded into the ants, but when they get together, they actually have a set of behaviors that they act like like in unison.

A

um Also I mentioned about self-care. Remember, there's some ants at the bottom of these uh of these things that are actually latching onto the grass and so, if you're feeling burnt out you're, probably one of those ants pop up for air and ask for help and ask for somebody to help you with with that task.

A

So um you also have another famous example, which are the uh which are the uh Cathedrals that you that you see here where you end up with a bunch of termites coming together, and they build up these really amazing structure, uh but specifically, there's a few properties that an emerging system has which is there's no Global control, there's not a single entity in the center. That's like guiding everything and saying you're going to do this and you're going to do this and you're going to do this and like mapping everything out.

A

There's also you couldn't even do it anyway, because there's no Global observability, you don't have the ability to see everything there are things going on in projects like I got a lot of visibility compared to many others, and I can tell you. I have just I can only see the tip of the iceberg. I cannot see everything that is going on, and so since there is no Global observability um that, despite that, we still have a convergence where that convergence occurs.

A

There's a given structure that that arises out of it, and so the system that has that structured behavior is not present in the original in individual components, and yet we still have this amazing beautiful thing that comes up, uh there's, also an emergent structure. So those are emergent behaviors, but you also have emergency structures where things get put together and welded, and so in a scenario you have these, uh these ice crystals that have formed and if you look at how water ends up working and how they all combine uh like you can have.

A

These type of structures arise and I also put this as a previous uh kubecon I believe it was 2019, uh Europe and uh I want to say Barcelona. But in the scenario we have a whole economy of people like you are in here: you're part of a you're part of an economy, you're part of a you're part of a organization of people who every group is acting in in different ways. And yet we still have all of these amazing things that come out that has Priyanka.

A

That's right before affects seven million other people that people are depending upon, and this is one of the central places where you can come together and be a part of that story and help influence the direction again, not control but to influence.

A

And so how do we influence? Then we start with Open Standards, so Open Standards drive commoditization, and we want to make it make sure we have the definition here. So when we say commoditization, we don't mean commodification, so commodification is where you're taking something that was like not sellable and you make it sellable in terms of commoditization.

A

What we're looking at from this definition is something that anyone can use that it's easy to access that you can just pick up and and start using, and so we want to focus specifically on Open Standards and specifically on the word open, so I know, there's a lot of definitions that people tend to have I tend to use the OSI definition of open and in terms of Open Source, it's something that is accessible to everyone, something that is approachable that like if I go and decide hey I want to go and use HTTP.

A

There's no one! There saying no, you can't use HTTP because it's a closed standard, and this is super important, because without these Open Standards are what Drive the ability for people to consume. But it also represents a place where we can, where we can influence, and so as we move forward that Open Standards ends up enabling emergence like we have IP and how much stuff is built on top of Ip. We have HTTP how much stuff is built on top of HTTP uh in the in Cloud native Community. We have things like the cube API.

A

How does the cube API end up affecting your ability to to deploy or to or to build products on top of so with the so with Open Standards like part of what we should be looking at is how do can we? How can we make sure that we're putting into the properties that we're influencing the structure and the making the structure predictable on what the outcomes are likely to be? Knowing that we're not going to control everything?

A

It's that we shouldn't even try to control everything, but we should try to put things in a way that are that are flexible enough, so that people can go and do interesting things, but we all converge back to a central to a central location.

A

So in and so in this scenario, we're saying Cloud native Embraces, Open Standards, and if you want to know what it means to be Cloud native I know we have the official term Cloud native and all the talking about, like tight, loops and uh and feedback loops and and similar.

A

um But I would also argue that if the standards were not open, if people were not free to come and use it, we would not be here. We'd be having a very small trade show at the most if we were lucky. So the reason why we are so successful is because we have been working on building new and even though they're not technically standards by the official definition like there's. No there's no board there in some of them, and actually some of them do have boards now.

A

But uh generally you start off with like well, let's go and and find something we can work on, that we can all come together and put a stake in the ground and say this is where we're going to start, it doesn't mean it gets ossified, but but it's something that in time we can continue to build and and rally around so, and so that means that, if we're embracing Open Standards and we're thinking about how do we set things up so that everyone can access?

A

So we get that diversity of of people like diversity of ideas. At that point, we are embracing the concept of emergence and embracing the concept that we can actually come together. We don't have to control everything so, like I mentioned, we are building standards like if you look at every single one of these things, even if you don't technically call it standards like there are protocols like you, have a kubernetes API that says like how can I go and deploy things.

A

I have like the oci that describes a variety of different things on how do I actually run and an image or how do I package a container image things like cni the control networking grpc so like you have all of these different things that people are rallying around and there are also opportunities that, if they're not going to be perfect, so part of it is these are some of the plays, not the only ones.

A

This is a very small list of everything that we're doing, but part of the challenge is, is learning what's out there so that you can then learn not only just what you're doing, but you can learn from others who you're working with, and you can tie it back into what you're doing for those for those standards and a really good example that I, like to use in a scenario is with like I.

A

Do a lot of work with spiffy and spiffy represents a path for how do we get identity where I've sped up my spiffy identity days for my workloads and another company might set up theirs? It gives the opportunity that I can consume their identities and I don't have to know anything about their internal infrastructure. We have something on the top in the same way that I don't have to know anything about your infrastructure.

A

If I'm using HTTP I can do a get call on it and it'll act in the way it's supposed to which might be give me something- or maybe it tells me, you're not authorized, but I'm going to get something. That's predictable. So with things like spiffy I'm able to get that identity across across boundaries and I can start reasoning about other other entities, and that has massive implications.

A

There's another thing to consider as well. It's not just about the technology. Like I said it's about the people and how many people here used to run infrastructure before you adopted kubernetes like you're the person who has to run the you're like a person is part of the Run book and they say: okay uh John. It's your turn to go and set up a script, John Amy!

A

It's your turn to go and uh and set up the firewall and something goes wrong, and so we have these really complex systems and again, thank you for Dusty for having this conversation with me, because this has helped me set this up, uh and so part of what we have to look at is what are the human processes that we have?

A

What are we changing, so everything that we put on our apis has an implication on what happens to in terms of our human processes, uh both positive and negative, because, like I mentioned, a lot of people have burnt themselves out trying to manage these systems because there's not enough of us here, but at the same time it also provides a path towards like I can build an image and I can give you an image, and if you want to run it, there's very there's nothing special.

A

You have to do and in order to understand how my image works, at least from the infrastructure side, like sure, from the application side, you have to learn something about how my image works, but that doesn't say anything about uh I. Don't have to give you very much information, because it's all in the specs, it's all in the standards, and we continue to improve on those so that everyone gains those uh those capabilities. So many of these processes have been standardized through the through the kubernetes API.

A

So please, like whenever you're thinking about standards or you're thinking about like what how do I build a protocol out always think of the people, I think about how people are going to consume it? How? How is this going to affect the people who are going to be up late at night, deploying deploying your code out um I want to also make a prediction as well. So today we now that we've covered immersions and how the how the entire space is an emergence system and we're an emergent Community. uh We can also think about.

A

Where are we going next? And so? If we take this as a as initial standard, one of the things one of the patterns that I'm saying is, we have a really clear definition on what an application is and that continues to become more generic and kubernetes is an orchestrator. It is definitely one of the most popular ones, it is becoming commoditized.

A

We have TurnKey solutions for it, but at the same time the applications themselves. They have different places where you may want to run so there's a there are teams over here who I've been working with, who have been working on 5G, Edge, Computing and they're, trying to work out how do I deploy my application so that actually lives close to you, like maybe you're, working in manufacturing and you're, trying to control a robot. The latency to your cloud is too much, and so you have something close by or self-driving. Cars are often another example.

A

You want something that's close by so that if you lose a link to your Cloud environment you're, you still have something operational, that's close by so so we have these applications that eventually we will start to schedule, have a manifest we'll have something we can describe. What are the? What are the requirements of the application then be able to deploy that application anywhere in the world based upon the properties that are there. If you need a GPU, you land you into an environment.

A

If it has a GPU, if you need, if you're a wasm based system, then we have to make sure you we land you into into that correct area and there's numerous places that today, you can start to deploy against. So not only when kubernetes and and clouds and Edge systems we're also seeing cdns actually start starting to run workloads in their actual environments.

A

The land literally like within some scenarios, within the kilometer of where you're at, depending on again, depending on your location, and so we have a whole set of protocols that have been serving us well here in the cloud native and I believe that we're able, if we are careful as a community and thoughtful with how this is how this works, I, believe that we can take it to the next level and set it so that, instead of just being bound primarily to kubernetes, we can actually make this so that it's something that is useful globally, that when my infrastructure, uh things I'm working on connects into your to your systems that we have those top level apis.

A

Those top level systems, of course, an escape hatch, because you sometimes need things that are not provided by those systems, but if we can get most of our environments to think in that specific way, with uniform approaches or close, uniform approaches to this and get down to better standardization I think we can have something that's much bigger than just kubernetes and Cloud native and just what's within this community. So with that, uh this is the main point I want to make.

A

Is that the standards that uh that we have that will continue to evolve, and they will allow us to to schedule anywhere to to have that to have that interaction where we can then predict like we can then deploy to anywhere or any particular group based upon the needs of the application rather than uh rather than what was the system. What infrastructure was that system designed for so with that?

A

I want to thank everyone for for your time and have have an amazing kubecon.