Innovations in Serverless Technology and AWS with Ramya Sarangarajan | KubeCon SLC 2024

This is Textron tv. Hey, everyone. We're back.

We're live here at CubeCon. If you caught our earlier attempt at going live, we apologize. Someone behind the, our booth here actually came in and accidentally stepped on the plug.

We're not gonna say it was intentional. Yep. But, um, so we're starting over from the top.

And if, if you've already seen the first five minutes, I apologize, but I think it's worth repeating. Anyway, we are live here at Salt Lake City at, uh, cube Con. It is an amazing cube con.

The buzz is palpable. There's a lot going on. I want to introduce you to our very first guest of the day here at CubeCon.

It's Roia Han Sarhan. No, you're close. You're close.

If someone pulls the plug again, the next Yes. The next time you'll get it. Yes.

But Rom is, is with AWS and we're gonna talk about her role here at AWS. But before we do, let's kind of recreate roia. I, I know I asked you this already, but it's been quite a opening here for Q Con, hasn't it?

Yeah, Absolutely, Alan. So this is my second cube con Mm-Hmm. I've, uh, I was in Paris, and, uh, the energy in Salt Lake City is, uh, is, is fantastic.

As I said, uh, one of the things I like about coming to these conferences is one, you get to interact with people like you from the industry who have a different viewpoint than, than than us who are sort of creating these products. But more importantly, we get engaged with our customers who are using our products on a daily basis instead of hearing what they like, what they don't like. More importantly, and these conversations are just, um, uh, so valuable to us.

And, uh, we cherish these. We make sure we have a good, large presence. We are able to talk to as many people as possible, and we make sure we take all that back to our leadership and our product teams to ensure that, uh, we are always building backwards from our customers.

So, yeah, it's been absolutely fantastic. And as you and I discussed, the weather's been great. Uh, it snowed yesterday, so, uh, yes.

I have nothing to complain about coming from sunny California. You know what? I, so I grew up in New York, but I live in Florida now.

I haven't, I'll be honest, I haven't seen it snowing in a long time. Ah, Yes, it Was good. It was good for about the first five minutes.

Yes. Then I remembered why I moved. Um, but, you know, there is a critical mass of people here, of course, in about two weeks, two and a half weeks.

Uh, AWS reinvent is time. Yes. And that's 65,000 people gathering in Las Vegas.

We'll also probably be cold. Yes. Um, but that's quite an event as Well.

Yes. And it's, uh, uh, I know my team's, uh, been so busy because we've, uh, been working on putting this show together. We are working really hard to ensure that we have a great show for our customers and partners at Reinvent.

So yeah, we are looking forward to that. Um, you know, we always, as I said, appreciate these opportunities to share our innovation and hear back and yeah. Reinvent is gonna be special as well.

And, uh, we hope to sort of see you and engage further with you there as well. Absolutely. We're looking forward to it.

So, Roya, before we jump into serverless and modernization and Kubernetes, I wanted you to tell your story. Mm-Hmm. A little bit.

Share with our audience. You know, there's a lot of people out here who say, I want a career in tech. I'd like to work for AWS, but I don't know.

Yeah. You know, I went to school, I went to school for technical stuff, or I went to school for business or marketing or what have you, but I just, how do you go there? Yes.

So if you wouldn't mind, share your story. Yes. Um, so I didn't grow up in product marketing, so, but my role today is, uh, leading product marketing for both Kubernetes and our serverless products.

So really the serverless container world for AWS. But, um, I, I start way back. I started coding when I was eight.

So I actually have lived with computer science and technology for the long, for the longest time. So I got my undergraduate degree in computer science, and then, uh, went on to actually do a sales job selling hardware and software. Really, that's because I was addicted to coding.

And so I, I said, okay, for my own sanity, this is not good because I will be coding all night Uhhuh. And so, um, you know, the sales piece was just so different from who I was as a person. I was a hardcore engineer.

And this sort of exposed me to so many different skills and, and, and, and, um, experiences, which is, was so invaluable because I was talking to customers day in and day out and post that, I said, okay, I need to learn more. Uh, and I went on to get my MBA and then went into management consulting because I was always in tech and I wanted to understand how healthcare works, how oil and gas works, how financial services work. But I would always drift back to like technology projects even then.

So then I decide, okay, I've done, I'm done with the experimentation. Let me go back to, uh, what is my sort of sweet spot. And I joined, uh, VMware in the mergers and acquisitions team.

Yeah. Uh, and then moved to product and, uh, product marketing. Um, and really, uh, two years back when, um, uh, when I had this opportunity to join AWS within the serverless business, I was super excited because, you know, uh, that's where, you know, a lot of changes were working.

And then about nine months back, uh, I also took on the Kubernetes job. So, uh, it's been a fascinating experience, uh, sort of knowing all sides of the business in a way, sales product and product marketing. But, uh, I enjoy this role where I'm able to sort of craft stories on our products for our customers, and also be able to, you know, communicate back to the service teams, the folks who are building the engineers on what our customers like and dislike about our product.

So yeah. I've vastly enjoyed this role. You know, what a great story though.

What a well-rounded resume. You, you come from that technical side. Mm-Hmm.

Did that went into sales. And you know, what they say about sales salespeople are born, not made. Yeah.

Right. But you aren't born to be a salesperson, but you, you still have that under your belt. Went into mergers and acquisitions, m and a real hardcore tech, you know, hard business, bringing it all together into product marketing for you.

And, you know, there's a lesson out there for people who are out there. Next time you hear someone whose title is VP of product, man, marketing manager, product marketing, don't assume that they're just marketing people or that there's anything wrong with just being a marketing person. Wonderful.

Most of these people have, you're not as unusual as a, as I as you may think. That is where these people come from. They, they have that technical background.

You can't do this job without understanding the technology. I'm sorry. Yeah.

Right. And so, a little bit of respect for those people. I wanna turn to something else you mentioned that you had of product marketing for serverless and Kubernetes containers, Kubernetes.

Now, they're not necessarily joined at the hip, though. They kind of are now, but they didn't, they weren't always Serverless was out there first. AWS and of course with Lambda probably put serverless on the map.

Yep. Right. But there are a lot of people in our audience who may be, they've heard, everyone's heard of serverless.

Mm-Hmm. Some may know what it is. Some may think they know what it is and some don't.

Why don't you explain serverless to them? Yes. So you're right, Alan, that we, AWS pioneered serverless about 10 years back.

Actually, we are celebrating the 10th anniversary of serverless as it is 10th, the 10th anniversary of Kubernetes as well. So it's almost like, for me, it's a big year of celebration for the businesses that, uh, that I sort of look after. But, uh, serverless 10 years back started off with Lambda really in a way where we said, how do we abstract the cloud even further?

You know, the CL cloud is born because we, uh, told customers that, you know, you should focus on what differentiates your business. And running data centers, managing the, the, the cooling and, and patching the OSS was not something that we said, Hey, you should be spending your time on, you should be spending your time on what differentiates you, what you're good at. We are a financial services company.

Do that because that's your core competence. And in a way, I serverless is the cloud and steroids. Right.

It sort of abstracts is even further allowing you to focus on what differentiates your business and business value versus, you know, managing the infrastructure underneath. So, uh, and you're right in that both Kubernetes serverless containers are the, are in this umbrella of modern applications, or what we say modernization technologies. Um, so if I think of serverless, we think there are four pillars which basically define it, right?

One is we abstract a complexity and how do we do that specifically when it comes to infrastructure management, right? Uh, where we say, okay, you don't have to deal with the complexity of managing your, your servers, your, um, your nodes. And we abstract it such that you can focus only on what's important.

We help you write as little code as possible. So we say minimizing coding is very important because, as we all know, have written code, code is a liability, right? It, it's sort of the more lines of code you have, the more more you have to deal with.

So how do we make sure that you build, you write as many as or as few lines of code? And how do you, how do you bring in more composability into the, into the picture. Number three, we say, how do you drive higher performance, resilience, availability?

And, and really a tentative serverless is AWS specifically with serverless, takes care of a whole bunch of the undifferentiated heavy lifting of these pieces, which is how do you drive high performance? How do you drive high availability, resilience? We call these well architected fundamentals.

We say, we take care of all of that so that you don't have to worry about it. And last, but not the least, and I think this is perhaps so, uh, it's one of the most interesting pieces of serverless, which is started with the event driven architectures piece, right? Where compute comes up with lambda when needed, goes down, when it's not necessary, but it also stretches to how you build your systems, how you architect your systems, because it's a sort of a bunch of microservices that talk to one another.

And we say architect for extensibility and integration. What that means is, uh, you're not just building your service for what you see today in terms of needs both articulated and unarticulated, but how do you build something that you can expand to pieces that you didn't anticipate five years hints. And so that way, it's great for that because think of it, I always say it's a highway system versus a sort of a suburban road system, because you, in way, can connect to the highway and you can add new capabilities and new pieces, and in fact, new scale if needed without having to restructure and re-architect your entire system.

So we think that's the four pillars of serverless. Yes. And, um, we see customers sort of take to it.

Some folks love it. We say there's a serverless first model. And one last point about serverless is it really started off with compute in Lambda, but today it's sort of extended to many other technologies, right?

Whether it's storage systems, our ai ML systems, our, you know, generative AI systems. And the reason is, uh, in our world, we believe that it's an operating model in a way is how you do things. How do people, processes and technology come together for you to accomplish more, get to market faster, drive agility, while at the same time not compromising on costs or resilience and availability.

So I think we, we fundamentally believe that there's a serverless operating model. Love it. So now, I said before all of a sudden, Kubernetes has kind of joined at the hip here.

When did that happen, and why? I, I think it's tied into the whole modernization thing. Yes.

But let's just look at Kubernetes and serverless and then we'll come to modernization. Yep. So, um, personally for me, it was when we thought of how are people thinking about their workloads, and how are we, how are folks thinking about, especially when companies have run workloads in their data centers and wanna modernize pieces, what are the technologies they're adopting?

Serverless being one. But of course, Kubernetes is almost becoming this sort of defacto standard from a platform engineering standpoint. And in a way, um, the way we think of our umbrella is there's Lambda, which is serverless, you know, reactive compute, event driven architectures.

We have ECS, which is our elastic container service, but that runs really well on a serverless container platform called Fargate. So that's sort of the world in serverless and Kubernetes is just a different way of doing things. So, uh, you know, I'm often asked, you know, what is, how do, how do these things merge?

How are we, how I, how do I look at the world? And I really say this, these are, these are different operating models. You look at the serverless operating model, and I call it the Kubernetes Operating Model.

And really with serverless, you know, you, especially in AWS you think AWS is the API. And with Kubernetes, you have the Kubernetes API, and you're engaging with that. And, uh, we see customers make these choices sometimes and often then we, we don't get into conversations where they're really debating one or the other, because if they wanna use Kubernetes, they've already made up their mind on various factors.

But, uh, we see the biggest drivers being skill sets, right? These require unique skill sets, the ecosystem of tools that customers engage with, interact with, right? Some of them are built on Kubernetes, some of them are built, say, on AWS.

And so you wanna engage with them and integrate with them heavily. Uh, it can also be dependent on various other factors, like what a client is using or what a customer is using. And you sort of just define your technology choice based on that.

So that's really the, the world where it comes in. And interestingly, many of our top customers use both the serverless and Kubernetes technologies. And so, uh, we find that they sort of, uh, we, we, they come in more often, but there's never, there's seldom a debate between the two.

Excellent. So you used the modernization word before the M word? Yes.

So when I first CloudFirst burst on the scene in 2005, 2006, everyone was talking about modernization. Mm-Hmm. But modernization to them was lift and shift.

Yes. They took their application from a data center, stuck it on a cloud instance, AWS was the only public cloud, and, and that's it. They modernize.

Yep. Today, when we talk about modernization, it's a very different playing field. We talk about making something cloud native, which oftentimes is Kubernetes and containers.

We talk about using, you know, a multi-threaded type of, uh, of, you know, architecture. It's, it's a, there's many ways to skin the modernization task. Yep.

If you talk to someone in the mainframe world Yeah. Modernizations means maybe separating the systems of engagement from the system of record, right. Or moving off of, uh, mainframe altogether.

Yes. Yep. That's also modernization.

Yep. However, today when talk, you know, a lot of people talk about modernization. We really kind of zero in on how Kubernetes helps with that.

Mm-Hmm. And, and quite frankly, serverless. Mm-Hmm.

But serverless is deemed to be a modern Yep. Architecture. Yep.

So let, let, why are people making that decision? Is it that you mentioned a little before, sometimes it's a vendor or it's a preconceived notion, but is it, it seems to be the overwhelming choice Yeah. For modernization.

Yeah. Why? I think, uh, even before I get into, it's actually a very interesting question.

Why are folks really thinking about modernization, right? What are the driving factors? It doesn't come at no cost.

No. You, you, you perhaps have to just remove away from what you're doing on a day-to-day basis and refactor things for a bit. Or rehost retire, as you said.

Uh, just lift and shift, right? Yep. These are pieces, and, and, and the motivations really to, to what you said, Alan differ a lot depending on the company we are dealing with, the company size, the industry, um, we, I can I, we still draw some common patterns.

Uh, number one being how am I able to increase my velocity of software development? I think that's been the, back in the day when I used to work on agile methodologies, and that was considered the modern way to build technologies, right? And we see that's one of the biggest sort of factors.

Number two is really the scalability angle, right? We spoke about how do I build systems which are not restricted by the infrastructure footprint I have, and how am I able to sort of scale, not just from a sheer demand perspective, and I'm able to, uh, tackle incoming demand, but am I able to change my system? Am I able to make it tackle other things?

Right? That's a big piece. Interestingly enough, a lot of people modernize for security and performance reasons, because they feel that they can leverage cloud technologies.

Especially when you look at AWS because we invest so much in scale and availability and security, you are taking off that security operations burden for yourself. And I know this is debatable, a lot of folks actually modernize for cost reasons, right? Uh, in a way, because one, we look at, we do the sticker price comparison of fa, you know, how much am I spending on, let's say, if I compare it to on-premises?

And I know, I know modernization is not just about moving your on-premises workloads to the cloud. It means many different things. But if I look at those specific workloads, you're spending a lot of time managing, uh, your infrastructure on the on, on premises and, uh, how, and then if something new comes in, really for you to refactor those cards sometimes are not taken into consideration.

So we have a lot of customers that are moving for that as well. So I think these are the motivations when it comes to, you know, technologies. Like, and then the next question is, what is modernization?

Right? And, you know, you and I discussed, I've stopped defining what a modern app is because that changes by the day. But I would, I can definitely say that there are certain patterns of modernizations that we see.

One is, you know, and, and I always joke, it's, it's basically how do you modernize your entire software, software supply chain in a way? Um, two, get these benefits of agility, better performance, better cost, uh, being able to sort of move quickly, uh, to meet the, again, articulated and un articulated needs of your customers. But really, the tenants are one modern technologies.

We now see that they're built of basically loosely coupled architectures. What that helps with is your, in your dev delivery teams can work on these independently, and you can change things around quickly. Number two is they always look at automated DevOps pipeline.

So you're able to sort of ship code faster. The velocity is much faster than before. And in this new age, uh, we've, we've gone about 10 minutes without speaking about generative ai.

Alan, that's a record. That's a record record, right? We have 50 very good.

But, but, uh, you know, using modern technologies like, you know, coding assistance where I don't know what have to write long lines of code, but something automatically generates, uh, it for me. So generative AI is sort of becoming center stage when writing code. So I would say this is sort of the, roughly the tenets of modern applications or modernization practices, but it's really a change of how you, um, how people, process and technologies come together.

Got it. And that's how we look at it, at least from the AWS world. I, I agree.

I agree. Let me ask you a question. I, I know you work for AWS but I'm gonna put you in a funny position.

What makes the AWS modernization on serverless and Kubernetes platform? The right choice Yeah. For our audience versus, I'm not naming names, but other competitors.

Yeah. Um, yes. And it, you know what, why, while we, we sort of have this community working with us in serverless, I, I say the biggest validation on these pieces is it's not you who's just singing the tune, but a whole bunch of people are, right?

Mm-Hmm. And we've sort of had this, these movements, I would say to a serverless. Yes.

And even Kubernetes, to a large extent, we were sort of, uh, a very big investor in the ecosystem, uh, and that, uh, healthy adoption. But I think, um, for us, what makes AWS stand, um, stand apart is, number one, we have, uh, a high focus on sort of making sure the fundamentals just work, right? Whether it's availability, whether it's security.

I've had a lot of customers tell me, we choose you because of security, because of the fact that you focus so much on that. Uh, we believe in sort of making sure our infrastructure and the services that are built on our infrastructure are highly performant. And so, even the way we architect our regions, the way we architect our individual services, the isolation pieces we provide, I think our are just one of the things we, we try and make sure that that's the highest standard possible, right?

That's number one. Number two, especially with Kubernetes, right? We don't wanna abstract away the benefits of Kubernetes, right?

Uh, I think the beauty of Kubernetes is the fact that it changes so quickly, but at the same time, the innovation is not sort of constrained by what AWS provides. So, um, and you're able to sort of operate within the ecosystem and the upstream tools that are there. Uh, I know a lot of companies which build their entire technology sets on Kubernetes, and this is why we say we are able to bring the best of both worlds, right?

AWS resilience, availability, security, and Kubernetes. And one last piece, um, you know, we speak about is the integrations that we provide, not just with our own services, right? Our 200 plus services, whether you want security tools, whether you want monitoring tools, whether you want observability tools, but also the fact that we have an ecosystem of partners who really plug into the AWS piece.

So, um, I would say that's what sets it a partner, especially with Kubernetes. We also give you options. You wanna run on AWS cloud create, if you wanna run on premises, we have what we call EKS ksa.

Yes. Right? Which, which helps you run, uh, on premises, Hey, you want something at the edge or closer to you, but you don't wanna ma you know, you don't wanna buy the hardware.

We have EKS on outpost, which again, give you, so that, that sort of selection of, uh, choices on where you wanna on stack, Well talk about modernization. That is right. You know?

Yes. And, and you doing it yourself versus using an AWS Are you gonna go build out the edge in the, on-prem and the cloud? Exactly.

That's big CapEx. Yep. And then compute choices.

I mean, again, we have the largest compute choices in the world now, especially with all the GPUs. Right? You already, you have, you free to do what you want.

And I think the fascinating thing about this, Alan, is it's not what, just not pieces that we anticipated our customers to build. It's what they built. And they surprise us in a way, and we are like, wow, this is fantastic, right?

When you're able to build something that we didn't initially perceive, uh, possible with our technology. So, uh, that's what is most gratifying. You know, that's one of the beauties of technology, is you put your technology out there Yeah.

And people take it in ways you didn't think of. Yep. And that's, that's kind of what builds this whole thing.

Yep. I know you've got reinvented in two weeks or two and a half weeks without getting you in trouble or me in trouble. Can you maybe just, just share with the audience, no one's watching.

Um, so what do, what do we, what's in the future, near term future, maybe a little longer term future for the serverless Kubernetes modernization? Yes. Um, track, um, let me talk about Kubernetes, and let me talk about serverless, though.

You'll see these things blending, right? And I spoke to you about the four pillars of, uh, serverless, which in a way, I think dictates what we do as AWS and it, I can extend this to, uh, Kubernetes in a way as well, right? Uh, so the four pillars of serverless, extending that to Kubernetes, um, our biggest sort of investments are in how do we abstract away operational complexities specifically with Kubernetes so that your operations people and your developers can focus on what's on different, what's differentiated for your business?

You know, it was funny when I first got introduced to the Kubernetes world, a actually Googled, Googled, what's a container? What's Kubernetes? When I was, when I knew I was getting the job back in, uh, back in, back in about four or five years back.

And, uh, one of the first statements I read was, Kubernetes is fantastic for developers because it makes the life of operators, um, nightmare, right? And, and I think one of the things we've been focused on for the last, you know, six years of us having EKS is how do we make this simpler and easier so that you can ship court faster, you can have production ready systems, you don't have to have to spend a lot of time and energy maintaining systems. So we're gonna absolutely focus on this piece.

The other element, uh, in a way that we are gonna focus on, and you've heard our CEO say this as well, is bringing the, you know, keeping the focus on developers. How do we ensure that our developer experience is uncompromised? How do we make sure that they're able to co-write code quickly, um, and, and, and really be able to sort of get the best of the infrastructure that we provide, best of the tooling that we provide without having to, uh, spend a lot of time and sort of just getting things done, right?

So, uh, and, and really only focus on innovation. So I would say those are primarily what we are focusing on. We relentlessly focus on our fundamentals, and, uh, you will, you'll get sick and tired of us hearing about our performance improvements, but we just had a couple release, and we'll continue doing that, especially on the Kubernetes, uh, side.

We've had recently, we had a, a couple of interesting open source, uh, launches, and we're gonna continue investing in the community because we believe that's fundamental to our learning. Yes. And in a way, uh, we are able to sort of drive more innovation, which is, I think, in the benefit of all of us who play in this ecosystem.

Tell you a quick story. The very first, I went to a DockerCon in Austin, I think it was 2016 Mm-Hmm. I came back, we were just launching what is today cloud native now.

Mm-Hmm. I was gonna call a dock or something, and I said, we can't use the name Docker. They own it.

We'll have to go with containers. I said, there was this other thing there from Google called, uh, Kubernetes. It'll never catch on.

It's too hard. Yeah. That's why I, I'm still working and not retired, I guess new.

But, um, one other quick area I want to touch on. I know we're over time, but you did mention ai Mm-Hmm. You can't get out of an interview without talking about it.

Yes. This one. How do you see either generative ai, a gen AI machine learning Mm-Hmm.

How are these all coming together to help shape AWS's offerings to make developers' lives better? Yeah. To help operators, to make Kubernetes, maybe dare I say, easier.

Yes. Um, so you've heard of the three ways or three layers in which we approach generative ai. And I had said many times, so I won't go over that again, but in a way, we have a lot of our customers running their models on Kubernetes.

Mm-Hmm. Uh, and that is fantastic because it gives you that scale. It gives you the resilience, it gives you that ecosystem of tooling, which is fabulous.

When you're looking at building models that are specifically, um, you know, which are sort of, let's say large scale, we also have customers fine tuning these models on EKS. And, uh, that is fantastic because then you get, you're able to apply the contextual piece, but leverage something that's already there. Of course, you have the choice of, you know, using, you using, you know, uh, Amazon bedrock, right.

In order to connect to these pieces. Um, there may be a time, and I don't know, I I, I, if I can see into this crystal ball, which will tell me the future, uh, as you said, I would've, it would be both. Uh, yes.

Uh, but, but in a way, we think, uh, the choice would remain with am with bedrock, right? We, we think folks, customers would demand choice and not be this, this one way of doing things. And so we will continue investing in that ecosystem and of the choices that customers have Specifically with Kubernetes, one, I think it'll continue getting better at sort of talking to the GPUs and being able to help you build these large scale models.

Uh, but I also think, we'll, we'll get very good at inference. We are getting better inference and customers are using, uh, a lot of EKS for the inferencing. Uh, and really this is about our ability as, as, as a service provider to connect to different models depending on what you need, and being able to quickly process that and deliver insights.

And that's where I think even my world, sort of my, the two pieces of my two in a way, the legs that I stand on Yeah. Sort of come together because we think, um, you know, serverless in a way, and that the way you operate would be also able to connect to these endpoints. I think it shines in integrating endpoints and in a way, uh, these pieces can come together for you to infer, infer whatever you need.

So, uh, well, as I said, I don't have the crystal ball, but I think, uh, it'll be one of those things where we think we will be right, uh, front and center leading the revolutions that our customers are, um, uh, sort of embarking on, and sort of in a way continuing to provide those choices, not just for our biggest customers, but also for those who are trying to, uh, experiment and do something new, which is what I think I, as I said, it's always fascinating to see customers build what you didn't even anticipate. So, yep. Love It.

Romney, I have to apologize. We kept you here way longer than we said. It was Always a pleasure, Adam.

Everything. It was worth it. Thank you So much.

I appreciate it. Thank, thank you so much. Thank you.

AWS Serverless and Kubernetes. Check it out. Get ready for, uh, reinvent in two weeks too.

Yes, absolutely. But first we've got a lot more here to do at Cube Con. We're live, we'll be back in a moment.

This is Alan Shimel. We're out.