On this episode of The Six Five – On The Road, hosts Daniel Newman and Patrick Moorhead welcome Nick Coult, GM at AWS Elastic Containers (ECS) for a conversation on how ECS has developed over the past year, what “AWS Serverless” means and the impact of generative AI on the serverless value proposition.
Their discussion covers:
- How AWS’ Elastic Container Service (ECS) has developed and how customer acceptance of an opinionated orchestrator has evolved over the last year
- The term ‘serverless’ is used differently by many, and how AWS defines ‘AWS Serverless’
- How the refactoring of legacy, monolithic applications turns out and how AWS Serverless helps
- GenAI’s impact on the serverless value proposition, especially from the user/operator experience point of view
Be sure to subscribe to The Six Five Webcast, so you never miss an episode.
Watch the video here:
Or Listen to the full audio here:
Disclaimer: The Six Five webcast is for information and entertainment purposes only. Over the course of this webcast, we may talk about companies that are publicly traded, and we may even reference that fact and their equity share price, but please do not take anything that we say as a recommendation about what you should do with your investment dollars. We are not investment advisors, and we ask that you do not treat us as such.
Patrick Moorhead: The Six Five is on the road in Las Vegas at AWS re:Invent 2023. We are jamming out in the video booth here, and hopefully you will tune in for all of our videos with myself and my good-looking co-host here, Daniel Newman. Daniel, how you doing?
Daniel Newman: Thanks for the kind words, Pat. Yeah, it’s great to be back here in Las Vegas at re:Invent. It is booming. I mean, we walked into this place, getting back here to this future ready sound booth where we’re doing all these podcasts, Pat, it was elbow to elbow and I think we know why. I mean, look, it’s been a huge year. If you are in the business of technology, you can’t have missed the generative AI boom. But let’s face it, the parts and pieces, the picks and axes, the plumbing to get this done, this is the week where we are going to hear directionally where so much of that stuff is going. Pat, I’m so excited to be here, part of the analyst program at the event, listening to the keynotes and talking to these great guests on the show.
Patrick Moorhead: I want to dive right into one of the key themes, big news here at the event, and that is essentially about serverless, right? There’s only three ways you can do an application. Well, actually four, if you look at super-duper legacy, you can virtualize it, you can stick it in a container, you can be serverless. With that, I’d like to introduce Nick, who’s the general manager of Serverless for AWS. Nick, welcome to the show.
Nick Coult: Thank you. Yeah, it’s really exciting to be here. This is, I think my fifth re:Invent now. It’s always a huge show, always tons of stuff going on. I think this year in that regard is not unique. Every year there’s really exciting things to talk about. Yeah, so I’m general manager of an Amazon Elastic Container service, and so I work generally in the area of containers and serverless and compute and applications, modern applications. ECS is a big service. It’s used by customers in just about every possible industry you can think of.
Patrick Moorhead: Yeah.
Nick Coult: Why do people like containers? Why do they serverless? The thing that sticks out in my mind and continues to be true is how fast you can move. The faster you can move, the faster you can go from concept to getting your idea into the hands of your users and your customers, the better you’re able to respond to market conditions, to things that are changing.
Especially right now, you look at the things that have happened in the last year, the economy change, inflation, all sorts of opportunities and challenges came out of that. Now, Gen AI, large language models, the way that people interact with technology is changing. That’s what’s to me, the core of what we’re trying to do in ECS is help our customers move faster, respond to the market, and that’s the kind of thing that we’ve been investing in the last year. That’s the kind of thing we’re going to continue to be investing in. I got some things I want to talk about today with you all in terms of what we’ve launched recently and what’s coming. Change is a constant, but the way that I think about it is that there’s things that don’t change and that thing that doesn’t change is what we’re trying to do for customers and how we’re trying to help them.
Daniel Newman: Yeah, and I appreciate that overview and introduction, Nick. We talked a minute backstage and you are the professor.
Patrick Moorhead: And we’re not. Actually…
Daniel Newman: And stood in front of the class. It was a good little banter that we had.
Nick Coult: Yeah.
Daniel Newman: But I can hear it when you’re starting to share your vision that that’s something that you’ve done and enjoyed over your career.
Nick Coult: Yeah.
Daniel Newman: I think you’re going to enjoy the opportunity this week to really continue to share with the world. You just started alluding to it, but maybe just do the double click here on the evolution. Last year, Pat and I spent time at this event having these conversations with some of your peers and others talked a lot about container services. Multi-cloud is a hot topic. Of course, as you alluded to, whether it’s companies trying to be more efficient, running their applications and workloads, trying to get FinOps right, how you run your apps, where you run it. These are big important challenges. Of course containers are one of the enablers to doing it more flexibly, optimizing your business. Talk a little bit though about the last year or so of development within ECS and how your customers are growing in acceptance of AWS, is a company that had a different coming of age as it pertained to multi-cloud and elastic container and services of the sort.
Nick Coult: Yeah, so containers has been a really interesting journey in the industry, because you look back to when containers first were introduced, probably 10, 12 years ago is when people started using containers for real.
Patrick Moorhead: Right.
Nick Coult: It was a new thing then. Now it’s actually a fairly mature technology. I would say if you look at people developing new applications, whether they’re doing it on-prem or in the cloud, vast majority are going to choose containers as the way that they package and deploy that software. One of the reasons is that it’s actually a very traditional and understandable way to run software. A container is actually based on technology that’s been around for a long time. That’s a good thing because it’s very familiar. There’s not a huge leap that you have to go from, from developing software on your own laptop to deploying software in the cloud at containers. It’s like a very, very familiar process. But the thing that we’re doing and the thing that we’re looking at is what are the parts of that process that we can take away from usage you don’t have to do. What’s the core essentials of the thing that you as a developer or an enterprise should worry about and what are the things that you shouldn’t have to worry about? We’re constantly evaluating that set of assumptions and asking ourselves and asking our customers, do you really have to do this? Is this something you really have to do? One example of that is the servers that you’re managing, right? When you went from data center to cloud, you didn’t have to manage the physical servers anymore. I’ve done that. I’ve bought servers and racked them and managed data centers. You didn’t have to do that, and that was such a huge step-up, but you still had the VMs, you still have the operating system. I got to pick which operating system I’m going to run. I got to make sure that it’s patched and up to date, I got to figure out how to scale my servers.
We thought to ourselves, is that really part of this core developer experience? Does that really have to be part of building and deploying software? The answer is no. It doesn’t have to be. That’s where serverless comes in, right? Now, a lot of people think of serverless as serverless functions, which was kind of the introduction of the category with Lambda, but AWS pioneered introduced Fargate in 2017, which allows you to run containers. Take a container image, any old container image, one that you could run on your laptop. Now you can run into the cloud, but you never have to go to that step of saying, “Okay, which server am I going to run this on? Which operating system am I going to use? How do I make sure that that operating system is patched up to date?” You just run your container, you go straight to running your container. That’s a really good example of the kind of innovation that we continue to invest in of asking ourselves, what do you have to do and what do you not have to do? If you look at the things that we’ve launched recently or the things that are coming, it’s just more of that. It’s like taking away the heavy lifting. Taking away the things that are just going to slow you down when it comes to trying to deliver and go to market quickly. We’ve seen, I would say, a really big shift to where if you look at industries like financial services, at some point in the past, financial services would’ve said, “Serverless, no way. We’re not going to do that. We got to have access to those servers because we’re processing a trillion dollars in transactions a year. We have to audit those servers, we’ve got to have access to those.” Now they say, “We don’t have access to the servers? Great. That’s one less thing that we have to worry about.”
Patrick Moorhead: Right.
Nick Coult: “That’s one less thing that’s no longer on our list of things that we have to audit that we have to present to our auditors. We don’t have to have teams of people managing those.” We’ve seen this pretty dramatic shift to where now in a lot of industries, serverless first is actually the mentality.
Patrick Moorhead: Yeah, it’s interesting. There’s a ton of monolithic apps and I like to find monolithic apps as what we did back in the early nineties before virtualization came on. Then I think we’ve extended the definition that says, “Hey, if it’s virtualized and on a containers, it’s a monolithic application, because the OS and the app have to be married somewhat, at least on the infrastructure.” How do you see it playing out? Is serverless for new applications? A leading question after what you said about the financial industry is it for new applications or do you see people refactoring these legacy or monolithic applications for Serverless and what are you doing to help them?
Nick Coult: I mean, the answer is all of the above actually. If you look at something like Lambda serverless functions, which was sort of the original serverless compute offering.
Patrick Moorhead: Yeah.
Nick Coult: That’s very well tailored towards what we call modern applications, which is loosely coupled small components that are loosely coupled together using event buses.
Patrick Moorhead: Right.
Nick Coult: There, yes, you’re going to be building something new, you’re going to target Lambda and you’re going to get all the wonderful benefits of Lambda when you do that. There’s a lot of stuff out there though that’s not modern. I don’t want to make it sound like that’s bad, that it’s not modern. There’s a lot of really interesting critical business functionality that was built by people who knew what they were doing and it’s battle testing.
Patrick Moorhead: It was leading edge then.
Nick Coult: It works, right? I mean, there’s no better proof that something works that you got it out there and it works. Especially when it’s mission-critical to go say, “Yeah, we’re going to go rewrite that whole thing from scratch.” “Okay, but you got to keep that old thing running. You can’t just shut it down.” This is where containers are really interesting technology in my mind because yes, they do allow you to do things like modern applications, microservices and all of that, which is great, but they also can run old stuff. You can take a monolithic application and package it in a container and run it as a container, and you can do that serverlessly on AWS with Fargate, right? We actually see that. We see people running legacy Windows applications, where it’s like, “We don’t even have the source for this anymore. The person who wrote it is retired, but it’s still critical for our business and it’s sitting on a server right now in our data center. We don’t want to be managing that anymore. Go run it with Fargate Windows.” Now it’s serverless. Now you’re no longer managing a server, you’re no longer patching a server. Yes, you can still go refactor it. You can still go modernize it. We’re continually investing in capabilities and developer experiences to make it easier to modernize an existing legacy. I don’t want to say legacy, but older application, that’s a monolith.
Patrick Moorhead: Right.
Nick Coult: But in some cases they may never do that. It might not be worth it. It does what it does. It works. You don’t have to, you can package it up, run as a container, still get a lot of the benefits of serverless, actually.
Daniel Newman: It sounds like this is something you’re seeing across all industries. You mentioned, for instance, the financial services.
Nick Coult: Yes.
Daniel Newman: Which, Pat, we always kind of say, “Look, there’s always the highly regulated, those industries are always the toughest. They’re the slowest to move. They’re always especially cloud related, they look at risk.”
Nick Coult: Yeah.
Daniel Newman: It doesn’t sound like this migration is limited though. It sounds like you’re seeing it across all industries and all applications.
Nick Coult: Absolutely. I mean, we mentioned financial services, government, public sector is another one now that’s very interested in serverless for similar reasons. Media and entertainment, online gaming, aerospace, travel, we got the list goes on and on. Pretty much any industry, if you ask yourself, “Can I run my application at serverless? Is serverless, right? For me,” what we’re seeing from our customers is, “Yes, you can and yes, it’s right for you.”
Patrick Moorhead: Yeah, I think that’s a great message. By the way, whenever you start with finance and government, I’m thinking high risk and they’re typically the last to make any major, major shifts. So yeah, it sounds like it’s becoming more mature. Every year that I’ve come here. I think this is my eighth year at re:Invent, and I was always amazed at how many things were announced at the same time.
Nick Coult: Mm-hmm.
Patrick Moorhead: One of the things as analysts, we like to look at as a, “Hey, is the vendor using its own services for its own stuff?” I’m curious, what are some of the AWS serverless applications or use a serverless architecture that the company is using out there and offering out to customers?
Nick Coult: Sure. Yeah. You’re asking about what would we say dogfooding, like using our own?
Patrick Moorhead: Well, just any AWS services that use serverless architecture out there.
Nick Coult: Uh-huh. Right.
Patrick Moorhead: You can talk about the offerings that you’re offering, but also are there any other maybe potential SaaS or PaaS types of offerings that use a serverless architecture?
Nick Coult: I mean, there’s a ton certainly in terms of customers running on AWS implementing on top of serverless, running a SaaS application on serverless is extremely popular. I don’t, off the top of my head, I don’t remember which ones I can’t mention by name and which that I can. I know we’ve actually got some customer sessions this week that we’re doing customers talking about stuff that they’ve implemented on Serverless. Amazon and AWS itself actually run a lot of stuff on ECS on containers and on Serverless for the same reasons that the customers do. But one of the really nice things about that, about the fact that Amazon and AWS run on ECS and run on serverless containers run so many things on those technologies is that you’re getting that battle testing at scale.
Patrick Moorhead: Right.
Nick Coult: You’re probably not going to be bigger than Amazon.com, right? But Amazon.com, there’s significant parts of that retail website experience that actually run on ECS.
Patrick Moorhead: Yeah, I would say great answer.
Daniel Newman: It’s a very useful data point. If you can run an operation that large that it probably would translate pretty successfully to companies of various shapes and sizes across different industries. Pat, I think this is the moment of the conversation in which we talk about generative AI.
Patrick Moorhead: I mean, it took us 20 minutes to get here. Can you believe it?
Daniel Newman: Yeah. I think it might be the longest we’ll go in our conversations this week where we don’t.
Nick Coult: I’m actually disappointed you didn’t ask about it sooner.
Daniel Newman: Yeah. I was actually saying that there’s kind of like this running joke where you could just dub over three people talking at a tech event, and just put the words AI, AI, AI, AI, and everybody’s just nodding. This is great. But we are seeing this sort of pivot, this transitional moment. This is a really important inflection in our industry. AI is going from this really neat media concept to something that is being implemented at scale across enterprises. I started this conversation saying that, “Look, this week isn’t just about new stacks and isn’t just about new GPUs, but this week is really about what the enterprise looks like.” The enterprise, what it looks like between applications monolithic. You’ve got new born on clouds, you’ve got Edge and all kinds of data in the ecosystem. Then of course you’ve got this need to tie together tech debt, legacy. I call it good, old things to new and to bring them up to speed.
Nick Coult: Yeah.
Daniel Newman: By the way, generative AI is going to make companies more productive. It’s going to have economic huge, multi-trillion dollar economic value according to studies that we’re looking at and the data that we’ve come up with. But the server serverless has a big impact on this. The architecture has a big impact on this. The ability for companies to put this together, deploy it quickly, get access to all their data build fabric. I’d love to know how you’re telling this story, Nick.
Nick Coult: Yeah.
Daniel Newman: Of Generative AI, serverless computing. How is this going to enable the enterprises, the company’s running on AWS to run faster?
Nick Coult: It’s a really interesting topic, and I could probably take another 20 minutes just on this.
Daniel Newman: Let’s do it in 10.
Nick Coult: First of all, the way that I think about this, and a lot of people think about this, is yes, this generative AI, large language models is new, but in some sense, and one level, it’s not new in that it doesn’t change what we’re trying to do. It doesn’t change what we’re trying to do, which is help our customers go to market faster, do more, be more efficient, take away the heavy lifting. In that sense, it’s not any different. It’s another tool in the toolbox. It’s a pretty different tool, pretty significant tool. I do think that it’s going to change what software is even at that level, how people think about what software is and what an application is going to change. It’s also going to change how software and applications are built. I mean, we’re already seeing that with things like Code Whisperer that allow a developer to leverage a large language model to go build code. But over time, it’s going to change the very idea of what these products even are that are being built on these things.
I do think that Serverless can have a huge role to play in that new definition of what software is. Modern applications are going to become even more important. The idea that in the past you treated an application as a big set of code that you wrote, and it’s a bunch of business logic that does some stuff, and it’s all the business logic of what it does is embedded in the code. It’s likely going to be in the future that it’s more about workflows, that these are going to be these sort of loosely coupled semi-autonomous agents based on AI and large language models that are interacting with each other. The idea of an application might go away in favor of jobs to be done and workflows that accomplish those jobs. Those workflows will be very fluid. They’ll adapt. They’ll self adapt even. If you look at how is that stuff all going to run, because you still need computers, you still need networks, you still need storage, you still need databases. Those things don’t go away in a world of gen AI. In fact, you need more of them, but how they interact is going to change and it’s going to be much more on demand. It’s going to be much more fluid in terms of how the interactions work and what types of interactions are needed. Serverless is ideally poised to serve that need. If you don’t know tomorrow how your application’s going to behave or your workflow is going to behave because there’s an AI behind the scenes that’s continually modifying it, the idea that you’re going to say, “Well, we need to have a bunch of clusters and we need to scale those clusters and patch those clusters, et cetera.” I don’t think that’s going to be the way that applications are built in the future. It’s going to be modern applications built on serverless.
Patrick Moorhead: Interesting. Gosh, the fractalization of applications actually sounds like it’s going to increase. Let’s put gen AI aside. I know it’s hard. We probably spend. I mean, Dan’s question was two minutes long.
Daniel Newman: Well, that’s just because I like to hear myself talk.
Patrick Moorhead: Well, I do too.
Daniel Newman: Huge fan.
Patrick Moorhead: But you got to live in the room too. That’s great. Gen AI aside, you are building and taking advantage of gen AI and you’re part of building it, but gen AI aside, what is the outlook for serverless look like, right? I used to do products in a prior role, and you’re making trade off between a more products and synergy and pulling them together to simplify, “Oh, let’s do everything.” That’s not a reality. What does the future look like for serverless?
Nick Coult: Our mission isn’t going to change, which is take away as much as we can to leave behind the essentials of what it takes to build, deploy, and operate software. That’s what it’s all about. How can we make it easier and easier and easier to build, deploy, and manage software. We’re willing to question along that sort of chain of things that you have to do. We’re questioning every assumption about do you really have to do this thing? You’re just going to see more and more of that.
Patrick Moorhead: Removing friction.
Nick Coult: Removing friction.
Patrick Moorhead: Ironically, this goes back to the day Amazon was created as a retailer, which was removing friction.
Nick Coult: Yes.
Patrick Moorhead: Then that move to AWS.
Nick Coult: Yep.
Patrick Moorhead: “Let’s remove friction. Let’s simplify for developers.
Nick Coult: Let’s take away the things that you don’t have to do.
Daniel Newman: Yeah.
Nick Coult: And question the assumptions,” right? You look at the invention of the cloud with AWS and before then, I mean, it’s kind of hard to remember this time of what was it like before cloud? But the idea that you were going to run all your business on someone else’s computers, are you crazy, right?
Patrick Moorhead: Oh, very much so and we all question, it wasn’t secure.
Nick Coult: Right. How do you trust that the people in that data center aren’t going to steal all your data, all these questions that people had that now, nobody worries about those things.
Patrick Moorhead: Yes, right.
Nick Coult: In fact, it’s the opposite of why are you going to manage your own server in your own data center?
Patrick Moorhead: Right.
Nick Coult: That journey of taking away the things that you don’t have to focus on, and we’ll take care of them for you, I don’t know exactly what those things are going to be. Nobody does, right? These things of all of it over time, but that mission isn’t going to change.
Daniel Newman: I think AWS has done a great job of delivering optionality, and I think that’s a big part of the narrative. It’s the flexibility. It’s kind of the choose your own adventure, whether it’s been your silicon strategy, your compute strategy, I mean your services. I always laugh at the end of every quarter when I look at the launch of new AWS services, it’s like, “Well, I mean.” But this is the whole thing about removing friction and bringing the meeting the customer where they are. That’s always been kind of the ideology. To your credit, I think it creates a massive list because customers need a lot of things. But Nick, at the same time, it’s part of the reason you’ve been able to gain such a significant market advantage in terms of the size of your deployments. The enterprise is using your technology and the amount of data that AWS has under management. I think the technology you and your group and Serverless are building has a really big role to play. I just want to say thank you so much for spending some time with us today, Nick.
Nick Coult: Yeah.
Daniel Newman: And sharing your vision and sharing a little bit about what you plan to talk about here this week at re:Invent.
Nick Coult: Yeah, it was super exciting to be here. I wish we could have more time, but yeah, it was great conversation. Thank you.
Daniel Newman: Well, the podcast fairies might be listening in, and they heard in here that you want more time with us, and we would love to have you back on sometime soon.
Nick Coult: For sure.
Daniel Newman: Thanks so much for joining us and everybody out there. Thanks so much for tuning in. It is Six Five on the Road here at AWS re:Invent in Las Vegas. Patrick and I in the AWS podcast booth.
Patrick Moorhead: Yes.
Daniel Newman: Having these conversations stay with us all week. There’s going to be more of these conversations to come and of course, subscribe and join us from all of the episodes of The Six Five. But for this show, we got to say goodbye now. See y’all later.