Cloud AWS re:Invent Coverage | Barry Cooks, AWS

Clips
News
More from Cloud AWS re:Invent Coverage

Barry Cooks

VP Kubernetes

AWS

Inside AWS’ evolving role in the expanding Kubernetes ecosystem

A key theme at this year’s re:Invent has been improved developer experience, with the general discourse tying into Kubernetes’ expanding role as the platform clocks 10 years. For its part, Amazon Web Services Inc. has largely contributed to Kubernetes’ growth, with developments such as EKS Auto Mode and Hybrid Nodes serving as testaments to that commitment to innovation.“When we look at this, it’s not about the product or the service; it’s not about the feature that you’re building. It’s about the end user,” said Barry Cooks (pictured), vice president of Kubernetes at AWS

play_circle_outline 10 year anniversary of Kubernetes

play_circle_outline VP of AWS Kubernetes, Barry Cooks

play_circle_outline Differentiation between ECS and EKS container services

play_circle_outline Introduction of EKS Auto Mode for managed components

play_circle_outline Launch of EKS Hybrid Nodes for on-prem integration

play_circle_outline Enhancing EKS Security: Shared Responsibility Model and Extended Kubernetes Upgrades Support

play_circle_outline Introduction of EKS Auto Mode for simplified Kubernetes deployment

Info
Transcript

Barry Cooks, AWS

Barry Cooks

VP Kubernetes AWS

Barry Cooks, the VP of AWS Kubernetes, discusses the 10 year anniversary of Kubernetes and the developments in the ecosystem, especially at AWS. Kubernetes has gained popularity for its strength and open-source nature, attracting a variety of players in the community. AWS offers ECS and EKS container services to cater to different user needs. EKS Auto Mode simplifies cluster management, while EKS Hybrid Nodes enable on-prem hardware to connect to EKS clusters. AWS aims to provide freedom to users in selecting tools and services that work on EKS. The focus is ... Read more

explore Keep Exploring

What has been the topic of discussion this year in relation to Kubernetes and what exciting things are happening in the ecosystem, particularly at AWS? add

What fascinated you about Kubernetes on AWS? add

What are the differences between ECS and EKS container services? add

What is the idea behind EKS Auto Mode and how does it aim to make users' lives easier, especially for those new to the Kubernetes community? add

What does connecting existing hardware in your data center to an EKS cluster managed by a third party allow you to do? add

What is the goal of EKS Auto Mode and how does it help with the burden of upgrades in the rapidly evolving open-source project Kubernetes? add

What were the key principles in the design and building of the Auto Mode product in the Kubernetes space? add

bolt Powered by CUBE AI

Barry Cooks, AWS

search

>> Hello. Welcome to theCUBE live from AWS re:Invent 2024. Live here in Las Vegas, having a great time with 50,000 of our closest friends plus. I think one of the things that's really been the topic this year has been the 10 year anniversary of Kubernetes. And I think again, to leverage onto that, there's so many exciting things going on in the ecosystem, especially here at AWS. And I'm lucky to be joined by Barry Cooks who's the VP of AWS Kubernetes, and that includes the EKS service. So welcome on board, Barry.

>> Yeah. Thank you for having me. I really appreciate it.

>> So I think help people understand a little bit about your job and what it entails. I mean, I know having been here what VP means, but help people understand what it means here.

>> Yeah. I mean, so AWS has a pretty unique culture, which I think is amazing, personally.

>> Peculiar.

>> Yeah, peculiar. Yes. We do like to say it that way, in fact. So we go very deep. So one of our requirements, regardless of level, in fact you and I were chatting about it just beforehand, it's like my pager hopefully won't go off. Everybody carries a pager. Everybody's expected to understand their content to the bottom level and help make things successful for our customers. It's part of what I love about the culture. So in my case, Kubernetes at AWS, whether that's on EC2 or using EKS, is in my space and I'm super excited to work with folks on that.

>> Yeah. So I think one of the things that people, and we just had KubeCon a couple of weeks ago, and there was just a lot of great stuff going on in the ecosystem. AWS was there. A lot of people were there. I think what really fascinated you about Kubernetes on AWS?

>> Yeah. I guess there's two things here that are really big. One is Kubernetes has taken the world by storm to some extent. It has done some pretty amazing things for its user base, and I was very excited about that. I started my career at the bottom of the stack, and I've worked my way up over time and ended up doing a lot of these higher level services now. And the power of Kubernetes really attracted me. I'd seen it at play. I've talked to enough customers to really hear about the strength that it provides. And the other piece of it is that it's open source. There's this really interesting dynamic and in the open source community, especially this one where you have such big players and small players and enterprise customers, you this mixing pot of everybody that's participating, which is really fun.

>> Yeah. Again, you make over 60,000 contributions so far this year. I looked it up earlier. So I think that's great showing because I don't think AWS gets enough credit for the contributions that are made back into open source, and I think that's key. But when you're building a Kubernetes product here at AWS, who are you thinking about? What is that persona that you're building for?

>> It's actually a really great question. When we look at this, it's not about the product, the service, it's not about your feature that you're building, it's about the end user. Who are they? What do they need? What's their day job? And what about it is hard right now that we could make better? That's how we tend to approach it. So for us, a lot of times you have to actually understand the organizational structure. How does this company operate, and what are those layers and who's in charge of what pieces? So for us, you have these operations teams and these teams are in this middle ground of this warring thing between developers who are trying to push code and change into these systems. And then you've got all the backend components running on systems, and they're in between those things. They're trying to help developers quickly evolve their thinking, innovate, get things deployed, and at the same time they're trying to monitor these things, scale them, provide security controls over them. So it's this really interesting dynamic that they're sitting in. Those operators, those are our primary customers.

>> Yeah. We look at that as being really focused on that platform engineering, and those teams are becoming... That was the hot topic actually. One of the hot topics other than AI at KubeCon. Do you see that as that persona really has evolved? A lot of them used to be wearing one singular hat, maybe they were the VM guy or they were the network person or the storage person. Do you see that evolving and you're helping bring them into this cloud native Kubernetes world?

>> Yeah, for sure. So we have a really broad set of customers. So we get to see people on this journey at many different phases. And it's interesting because there is so much that they need to drive, and Kubernetes allows them to do this in ways that offload some of the mental pain of trying to keep track of things. So a lot of our focus is like, how do we help you operationalize the things that are really important to your team's success? And how do we help your developers get through that pipeline, get into production, get things deployed, and then keep eye on things, make sure they're working the way you expect it?

>> So EKS not the only container service in town.

>> It's not.

>> How do you differentiate between the two container services with ECS versus EKS?

>> Yeah. It's really easy for us. So in a lot of ways, one of the great things about AWS is that we have a mental model that says we are going to allow individual groups to target the use cases, the users, the people that they think are most important to them and be successful that way. And that's why we have both an ECS and an EKS. If you look at ECS, it's been an amazing platform for core AWS users who love the way our APIs work, the consistency. They jump into that and they love it. They eat it up all day long. If you look at what Kubernetes has done, it's this open solution. So a lot of the customers who are coming in the Kubernetes door are saying, "I want this standardized open API model." This I can use on-prem, I can use in AWS. Maybe it's some other cloud out there, I might want to use it on as well. It gives me a simplified way to think about things at a global scale.

>> Yeah. I think that is a great way to break it down. When organizations are looking at Kubernetes and looking at EKS, what are the positives and what are the things they should really take into account before moving to something like an EKS?

>> Yeah. There's a couple of different ways that people look at the cautionary side of jumping into Kubernetes. In the one hand, we have very sophisticated customers with large operations teams full of developers who know Kubernetes deeply. For them, I have to release control. I'm ceding some of this to AWS. I'm trusting you to manage components that I know really well. I got to be comfortable with that. I got to understand how this shared responsibility model is going to lay out and how that's going to work. And sometimes that takes a little back and forth with us. And then on the other side of the spectrum, you have a lot of folks who have heard about Kubernetes. It is very popular. It is in fashion, if you will. And they're looking at this going, "I don't have anybody with Kubernetes in their title. How do I take advantage of this technology? I know it's great, and how do I use that?" So for them, it's really understanding what are the key things that you are taking on your side when you go on the Kubernetes journey and how do we make that easier for you?

>> Yeah. I think that's a great bridge into my next question, which is Kubernetes has a lot of knobs. I mean, we've both had to dive deep on these things and understand. And when you start to look at it, you had some big announcements this week, and again, even on the main stage that Matt did a good job of going through some of them and some of the other things that were earlier on in the week and things of that nature. We actually talked about it with him in the analyst sessions and stuff. So help us understand what people who may not have heard them or missed them, what were the big announcements?

>> Yeah. So I think from our perspective, we're constantly pushing forward with how do we make your life a little bit easier? So we were talking about when you look at these users, what are they worried about? So one of the things with these people who are just joining that Kubernetes community, they're really worried about just the knobs, understanding all the components. How do I actually run this and what pieces can you take over? Even our largest customers with huge teams of Kubernetes experts, over the years, they've even come back and said, "I don't need to run this anymore. I've got so many interesting things I want to go solve in my platform that I could really use you taking over more of this responsibility." That's what we stepped into this year. So the first big thing that we had was EKS Auto Mode. And the idea behind EKS Auto Mode is we should just be managing more of this for you, if that's what you want. If you want to manage these other pieces yourself, we still allow that. We're not changing, it's just a new feature in EKS. But what it'll do is say like, we can stand up this cluster, give you a bunch of managed components in the cluster that we'll just operationalize for you. Key amongst these, everybody's favorite, how do I deal with CVEs and patching? So we've introduced a new mode of operating EC2 instances inside of EKS now, where we call them managed instances, "We will manage the EC2 instance on your behalf." That means we'll decide how many you need. You can give us some guidance, but we'll decide on how big you need. We'll auto scale them up and down based on your actual workload as it's changing. So we can do these things and we'll patch them. So now you can take advantage of our expertise in patching and rolling things out worldwide to make sure that your CVEs are patched as fast as possible and in a non disruptive way. So that plus all these other components of Kubernetes that people are used to installing and managing independently, we're now going to take over that as well. We'll install them, we'll manage them, we'll upgrade them, we'll scale them for you, and it just makes the whole situation significantly easier. So the bar to get into Kubernetes just dropped. It's much, much easier to get in now on that ground floor and really run workloads.

>> Yeah. I read through it and it was talking about another one that's difficult for a lot of people is the networking aspect, and they can see that to you as well, and some of the storage aspect. What do you see as, again, people are most excited about? Because you had another announcement and you talked about it. A lot of people go to Kubernetes because the concept of portability of containers, but maybe they want to have some that are on-prem, and you have a new announcement around that as well.

>> We do. So if you look at Gartner numbers and others, they'll tell you that there's huge interest in Kubernetes. There's huge interest in containers, and there's huge amounts of stuff still sitting on-prem that hasn't gotten through that journey yet. So as we've talked to customers, we've talked about what would be the best way to help you be more successful at moving from those on-prem workloads and for various reasons into more of a modern architecture. So we launched what we call EKS Hybrid Nodes, and what this will allow you to do now is take existing hardware in your data center, connect it into an EKS cluster that we run and manage. Maybe you're in Auto Mode and we're going to take on a whole bunch of that operational structure for you. You can now just point containers at this on-prem component, have it be a node in your cluster. You can have nodes here, nodes there. It really helps you to start the journey in stepwise fashion. We have a lot of customers who are trying to exit data centers. Data centers take a while to get out of. CapEx cycles are seven years in many cases these days. So if you've put a bunch of money into a whole slew of hardware, you want to sweat the asset for a while and this will let you continue to operate those assets until you're ready to age them out of the system and then go full cloud.

>> Yeah. I think to me, that was the really interesting part of it, it was again, being able to have the control plane and how it's stretched Kubernetes control plane say that 10 times faster. But I think what is a lot of fun is that there's so many different projects around Kubernetes in that ecosystem. How do you see organizations really playing? Because to me, to your point on the operator's, part of it is sometimes they want to use the service mesh from this, or this other piece from here. How much levels of freedom do they have with EKS for that stuff?

>> Total freedom. So our view is we don't want to dictate answers to people. I was having a conversation with some partners actually here. I'm like, we're not in the business of kingmaking. We love partners. We love all of the partners. We like the open source ecosystem. And our view is if you like it and it works on Kubernetes, it works on EKS. We're not restricting anyone in those cases. And we think that is the best way for customers to say, "This is the piece I like and I have my own reasons why." And we have people who want to run Istio. We have people who want to run other things, whether they VPC lattice in some cases, we have things that we offer as choices. The open source container ecosystem has many, many different choices, and we love them all. Because for us, those are just different ways to solve problems for customers and we want the customer to decide the best for them, not us.

>> And it would seem like a lot of these large and small organizations, they may have relationships with security vendors maybe that for instance, that they're using in other parts of, is that another place where security comes in and there's stuff that you take on versus that shared responsibility model?

>> Yeah, it is. So with EKS Auto Mode, we're trying to take on more of that burden as much as we can. We're also trying to take on a lot of the upgrade burden. So if you're at all familiar with Kubernetes, the funny thing about Kubernetes is it's an open source project. It tends to move pretty quickly. It evolves its API model over time, and that's generally super valuable. It revs multiple times per year. And then you go look at the people who are trying to take full advantage of it. They're huge enterprises in many cases, and they love upgrading their systems pretty much never. So you have this distinction between the two of like, I can release this thing three, four times a year and I don't ever want to touch my thing. It seems to be working right now. So we're trying to help people bridge that gap. We did a launch earlier this year of extended support for Kubernetes. The reason that we did that is because what you don't want to do is fall off the back end of the open source train and have no support for CVEs. No security support, now you're flying blind and you're taking huge risk. So by introducing extent support, what we did is we said, "Hey, take your time, upgrade on your cycle and we will back you up with that." And that allowed these customers to say, "This is now my way to go forward because I can take the amount of time I need, and I can work on getting more current over time, but I'm not being forced functioned into it by threat."

>> They may have a N+2 or something like that type of where it's released +2 or 2 versions or take every other patch or what have patch roll-ups. That had to be huge. I mean, for customers especially, I know regulated industries and others where maybe they can't go as fast because they have to... The workloads that are on there. What workloads are you seeing come on to EKS?

>> That's a great question. I mean, we have enough breadth. We're running tens of millions of clusters at the moment, so we have... It makes me a little nervous, but we have huge scale. So the great thing about that is you see everything. If it were a popularity contest, GenAI, hands down. Everybody knows that that's super popular. We are an amazing platform to GenAI training on. We see a lot of very large workloads in that space. It pushes us in really interesting ways, but we see financial services as a huge area. Government and other regulated industries are big. We see lots of classical web services. We have a ton of gaming customers who have low latency edge connections. So you end up seeing this huge mix of workload types. And one of the nice things about that is, for example, what we love to do is take a customer and another customer who is in a similar space and put them together to talk. It's one thing to talk to me and I can tell you all the things I think we're doing that are awesome, which I love doing. It's another thing entirely for a gaming company to meet another gaming company that maybe is from across the world here at re:Invent, sit down in a conference room and say, "Well, what have you seen when you tried to do this?" And we do this a lot. We actually try to get our customers together and we basically are moderators at that point. And they're really in there digging into the details of how they're trying to solve things. And we learn a ton just from those experiences.

>> Yeah, I know. I was out to dinner with one of your customers who was also doing it for another customer in another industry, but they had the same security and data concerns about things of that nature and scale concern. When you look at EKS and somebody's coming in, because we hear this all the time with GenAI that just the scale is massive. What are people thinking about from where they can start to where they can get to with EKS?

>> I mean, right now we don't think we're limiting people on scale. I think that the problem with scalability is anytime you want to talk about it, it's like an asterisk and under that is it depends what you're doing. So that's true in this world. With these variable workload types, they put different pressures on the system. We continue to scale with our customers. We sit down with them every week and we talk to different customers and we say, "Are you hitting limits? Is there something we can do to make your life simpler? What are the kinds of limits?" We have customers who will say, ""I'm going to deploy one pod on one machine, and that's going to be the biggest possible machine I can get, and I would like to get 50,000 of those machines." And we have other customers who are like, "I have 50,000 pods and I don't need a lot of machines. And they're not super compute intensive, but they need a lot of networking." So you end up seeing a lot of these mixes, and we try to stay one step ahead of where our customers are at. We work hard to earn their trust so that they know if they come to us and they say, "I'm going to need this next year," we are going to give it to them next year.

>> Yeah. I think again, it's such a fast evolving. We were joking at KubeCon that was talking to Taylor from the CNCF and we were having this conversation and actually the upgrade thing came up. So I'm glad you hit on that because that's the one thing that everybody agreed on. Even 10 years in, upgrades are still a pain in the butt. So it's like you start to look at it. When you start to look at out into the future and things like that, obviously there's the open source portion of it and then there's the AWS portion of it. How do you see the two sides evolving and what you guys are involved with?

>> Yeah. It's a bit of a balancing act to be honest. I mean, we're running a business and we try to sell value to our customers. And at the same time, we have a moniker, if you will, that we talk about internally all the time. And our saying inside the group is, Kubernetes is bigger than us. If it stops existing, we don't have a business. So it's super important to us to foster a healthy community. It's also really arrogant to assume all the answers yourself, and you don't need those communities. I mean, I talked about how much time we love spending with customers because we learn in-depth how they really do things, and I think the community has the same function. It's really this meeting of the minds around what are the problems you're seeing? What are the problems we're seeing? How can we collectively solve these in a way that's going to actually move the whole group forward? It's that game theory thing.

>> Yeah. And it would seem like, again, because I think that's what when I looked at Auto Mode, I looked at it and I'm like, okay, great. So this is almost a bridge from people who were maybe ECS where there's almost no knobs there. You go through Fargate and you set it up real easy. It's serverless. Often different infrastructure to here where I have some of the knobs that I may want to turn, I may want you to turn some of these knobs. I may want to turn some of these knobs. This becomes that flywheel that helps people get up and running faster on those clusters.

>> And that's exactly the idea is, is when we looked at this, we said, "There's some amazing things we've done in the serverless space and how could we translate some of that joy to the Kubernetes space?" And it really boiled down to take a few of these things off the table, but give them the choice. Give them more flexibility than what we would do in a fully opinionated solution because that's what this community is used to. They like choice. So one of the keys in Auto Mode is we're still fully Kubernetes conformant. You can still bring in and run anything you want out of the CNCF ecosystem that works on Kubernetes, and that was a really important first principle in the design and building of that product.

>> Yeah. I think that's key. Again, I think I know the answer to this next question, but I'm going to ask it anyways.

>> Go for it.

>> I think what are some of the popular, I mean, you hit on it a little bit, popular verticals or industries that you see really navigating or navigating their way Kubernetes?

>> Yes, exactly. Grab the wheel.

>> I try to not go down that path, but it was like it hit me there, but what are the industries and verticals that you see playing into this?

>> Yeah. I mean, I think we see financial services as they oftentimes open source. They like being able to drive. They work at huge scale. So we work with a lot of financial services companies. That's been big. I mentioned gaming. When I got here, that was not the one I expected to spend a lot of my time with. I do spend a lot of time with gaming companies who are doing interesting things at the edge. So they have these interesting combinations of, I need to go reach players in the city at a very low latency, and I have other tasks that I'm going to do on a different frequency, and they want this all interconnected. That's been a use case that we see quite a bit. And then the classic web scale services and people who are running web-based applications is of course bread and butter.

>> Yeah. It would seem that way. In fact, I was having a conversation with another end user customer, and one of the things that they were asking me, they asked me, they go, "Well, should I go bare metal or should I go Kubernetes for AI?" And I said, "Well, are you going to compete with ChatGPT or one of these other services, or are you looking to build your business?" They're like, "We're building our business." Well, it seems like, from all the people I've talked to, most GenAIs actually happening on Kubernetes on services like EKS. Are you seeing that where this is people who are looking for a simplified way to scale out and contract because as Swami was saying, not doing it all the time, and we need to contract those services automatically. Is that what you're seeing as well?

>> Absolutely. So one of our most popular use cases in the last 18 months as things have really taken off in this space, there's tons of experimentation, there's tons of exploration, and there's tons of scale up, scale down. So you have these batch style training jobs that need to be able to scale up to a very, very large set of infrastructure. You do not want to keep running that very, very large set idle. So things like Carpenter, which is a open source project we developed, donated it to CNCF earlier this year. We were super excited to do that. We've been really excited to see other cloud players joining that community and helping to drive it forward. It does that. It helps you to really go in and scale up, scale down, lay out your infrastructure the way you need it and make those decisions. And we see most people coming in the Kubernetes door.

>> By the way, loved it and I loved the name of it too, because I thought it hits the nail on the head.

>> Yes, exactly.

>> So as we look out into the future, Kubernetes is 10 years old, looking out the next year to 10 years, what do you see on the horizon for EKS?

>> That's a great question. We spend a lot of our time debating this one, as you can imagine. So I think we talked about platform operators and what do they need and how do they operate day to day, and what are the challenges they still face? So things like EKS Auto Mode we think are a huge help. We launched Crow at KubeCon actually recently. We also launched a Cedar controller at KubeCon. So ways to implement policy and security in a much more seamless way for your user base if you're an operator. Crow, ways to hand developers clusters with API endpoints that are developed and set up in a really consistent way. So we see ourselves moving up in these stacks saying there's all these other places on top of Kubernetes that could use the same simplification. So that's where I think we're focusing our attention is as we continue to move up and look at these things, there's a lot of work we see there that in conversations with our customers, they would love to see us go help them implement these pieces. And at the same time, never lose sight of availability, never lose sight of scalability, never lose sight of security. It's critical to our success, because you only get a fall down once on those.

>> Yeah. You stub your toe, that's it. And I think that's why people come to an AWS because you have the availability, you have the consistent storage. Are you seeing workloads, and I was when I was talking to Nick earlier around the ECS workloads, what kind of workloads and not necessarily industry specific, are they more ephemeral or are they more long-running type stateful type of-

>> We do see a mix of things, and I think if you look Kubernetes historically, very much ephemeral, very much stateless workloads, beautiful system for doing that. And because it works so well, people are trying to jam more stuff into it. They want to try and use other things because they love the benefits that it provides. So we are seeing a lot more stateful workloads, and one of the great things that we've done is try to help you connect into Amazon services. Because if you can connect into Amazon services, you can push some of that state into these well maintained, well operated services behind the scenes. So AWS Controllers for Kubernetes is an example that Crow, CRO or KRO, sorry. Of course, that's okay. Another example of that. And in these cases, what we're trying to do is help you with that other class of workloads that you really want to bring into Kubernetes because you want to get all the advantages that Kubernetes provides. But what you don't want to do is run a giant database inside your Kubernetes cluster and then have to deal with the stateful set that contains and having to understand how to drain that correctly or run multiple versions of it and synchronize it. What you want to do is come in and say, "Hey. I'm just going to use RDS and here's all the things and that's who I'm going to go talk to." And we try to make that as simple as possible. So you take full advantage of AWS services.

>> So they don't have to necessarily put a stateful database into EKS? They can, but they don't have to.

>> Exactly. It's back to that choice. We want to say, "You've got some open source database and you want to just run it because you love it, go for it. You want to take advantage of AWS services, we're going to make that super easy as well."

>> Yeah. And I would assume that you're, again, I think perfect timing with the hybrid, again, aspect of this and being able to go and have the clusters with nodes on-prem, given that some people have changed their licensing models and people are looking to monetize their applications, this may give them a chance to do that. Are you seeing-

>> It's very subtle.

>> Well, I mean, again, it's down the hall kind of thing.

>> No, we do. Absolutely. So at modernization, there are some announces at the show on Monday, in fact-

>> Q Developer stuff. I was very keen in on that.

>> Yeah. Look, we really want to help users get to where they want to be. If where they want to be, for example, EVS we just launched, that happens to be in my group as well. If you want to continue to run your VMware workloads in a VCF VMware environment, we now have a native service for you to go do that. And we're super proud of it, and we're very excited about that launch. If you want to go and say, "I've decided I want to use containers in a more modern approach, and I really like Kubernetes and I want to move into that model," we are going to support that migration and that move. And that is the classic on-prem to cloud move where people are on these journeys to modernize old applications. We run into people running COBOL and various other things, and we help them go step-by-step into a modern architecture like Kubernetes.

>> Yeah, no, I was talking about this with another customer as well, where again, they have those multiple tiers of applications where they're running Kubernetes. Then behind that, they have VMs. Those VMs are talking to a database that's on a mainframe. And again, it's been running there for 30 years.

>> I actually talked to one today that has exactly that architecture.

>> It may actually be the same customers. I'm not going to go down that path. Are you seeing that there's more... To me, it seems like app modernization, it cooled down last year but then this year, it's on fire again. Is that a lot of the conversations you're seeing as well?

>> I agree with you. It actually did cool off a bit last year. It was such, I don't want to say it was hyped, but it was such a focus of attention for folks. And I wonder to some extent if the GenAI stuff distracted people from that a little bit. But this year, yes, we are back in a lot of app mod conversations. A lot of discussions with our partners about how can you and I go together to help people move forward?

>> Yeah. I could imagine that because I think, again, we're at the tip of the spear with GenAI. I mean, I agree with Matt with what a lot, Matt Garman when he was on stage yesterday talking about how apps are going to be apps and GenAI is going to be part of those apps. These agents are going to be part of those apps. They're not going to necessarily be the app going forward, which I think lends itself to a Kubernetes environment where you have these services in different containers, much like Amazon develops its underlying infrastructure where you have different services that can be upgraded at different stages. Are you seeing that? That flexibility of architecture is also when people are trying to modernize where, "Hey, I want to have my GenAI running over in here and I'm going to have the database over in here. Or maybe I have a cache over in this part, and then the web front end and controllers and stuff like that?" And that's why they go down this direction.

>> It is. And I think to some extent, you got to be careful in this because you can get yourself in trouble by just dividing things into hundreds of ways without really having a thoughtful process behind it. But that ability to segment things and segregate them and put them out in a way that actually gives you the real value that you couldn't have gotten in this old monolithic style is one of the key factors that gets people excited about moving into this model.

>> So last word on this, where should people go if they want to learn more about EKS? I know where it's on the website.

>> Yes, of course.

>> You have tons of resources out there, so help people understand.

>> We do. So we have an amazing field organization that backs us up. They contribute enormous amount of open source. We have blueprints to help you get started. If you want to go do stateful workloads, we have blueprints around how to do that. We have that across a whole spectrum of different deployment models, different types of open source that we will show you how to do. And we will just hand you example code. So there's a whole bunch of that, that we've put out there. And we also will have specialists that are more than happy to jump in and have a detailed conversation about your exact problem, your exact implementation, and give you some advice on that as well.

>> Yeah. And I'm sure we'll see your team out at KubeCon in London April Fool's day, whoever decided to start it on April 1st, whatever.

>> We'll come up with something for it.

>> Come up with something. It'll be a good fun day. Again, Barry, thank you for coming on board. I think this has been a lot of fun, and I think really informative, especially about the new stuff and where you're going, making things simpler and helping people actually do things. This has been really cool. I really appreciate it.

>> Yeah. It's been my pleasure. Really fun.

>> So thank you for watching this episode of theCube here at re:Invent 2024 live from Las Vegas. Stay tuned.