Episode #122: Live from AWS re:Invent 2021 with Ajay Nair and Talia Nassi

December 6, 2021 • 52 minutes

On this episode, Jeremy and Rebecca chat with Ajay Nair and Talia Nassi live from AWS re:Invent about the latest serverless announcements, what serverless "phase 2" looks like now that people have accepted the operational model, the developer experience improvements being made with things like SAM Accelerate, and much more.

Ajay Nair is the General Manager (AWS Lambda Experience) at AWS. Ajay is one of the founding members of the AWS Lambda team, in his current role, drives the serverless product strategy and leads a talented team driving the product roadmap, feature delivery, and business results. Throughout his career, Ajay has focused on building and helping developers build large scale distributed systems, with deep expertise in cloud native application platforms, big data systems, and streamlining development experiences. He is also a co-author of Serverless Architectures on AWS, which teaches you how to design, secure, and manage serverless backend APIs for web and mobile applications on the AWS platform.

Talia Nassi is a Senior Developer Advocate at AWS Serverless and an international keynote speaker who delivers content on all things testing and quality. Previously, she worked at Split Software as a developer advocate and at WeWork as an engineer, and implemented Testing in Production from start to finish! She is passionate about feature flagging, canary launches, CI/CD, testing in production, and A/B testing. She has spoken at countless conferences internationally, ranging from audiences of 100 to 4000!


Jeremy: Hey everybody. I'm Jeremy Daly.

Rebecca: And I'm Rebecca Marshburn.

Jeremy: And this is Serverless Chats live from re:Invent, pretty exciting. And we can't really hide the guest. So this is our guest, our guest today and I'm going to bring up-

Rebecca: Is Rebecca.

Jeremy: Is Rebecca. And also Ajay Nair, who is the general manager of AWS Lambda functions or sorry, AWS Lambda Experience at AWS.

Ajay: Yes.

Jeremy: Thank you for being here.

Ajay: Always good to be back Jeremy.

Jeremy: Oh. Oh, we are broadcasting in the house.

Ajay: Yes.

Rebecca: You want to bet.

Jeremy: Going out everywhere?

Ajay: Yes. I always like to also be known by my alternative names wherever possible. Yes.

Jeremy: Right. Well, we are so happy you're here. Happy to be back in re:Invent. And I want to say just us for all the people that put this event on, AWS. Such an amazing job, the COVID protocols, everything that has been hap just to make this happen, make this event. Everybody's vaccinated, everybody's wearing masks. And please PSA quickly, go get vaccinated so that we can do more of these things. So that we can get through this and do that. So anyways, so we are super happy to have you here. And what I'd love to do just to start off maybe is just in case people don't know who you are, just give a quick overview. What do you do at AWS?

Ajay: So Jeremy, I've been with AWS now about close to eight years. I was one of the original launch members of the AWS Lambda Service. So over the years have worn different hats. So I started out as just the PM for Lambda, then became the PM manager and director for Lambda. Over the last year or so I've been running product and engineering for our developers, administrators, and operators experience serverless and building on Lambda in general. So, most of the things that you like and don't like about Lambda and blame on me in some form of fashion.

Jeremy: Yeah. Well, there are a lot of things we like about Lambda. So why don't we start off? What are some of the big launches that happened this week? And then maybe we can put those into context.

Ajay: Yeah. No, that's a great place for us to start off. Like you said, it's great to be back in a forum interacting with the community and customers and just hearing live feedback about some of these things. So we've had a collection of launches, at least the ones that made the re:Invent forum. I know we try to make sure that there's a cadence going through the year, like we had pre:Invent and pretty sure I'm sure there's going to be an official post invent-

Jeremy: Yeah.

Ajay: ... season as well. So, you should not count these as the end of them all. So let's see some of the things we had this week. So we talked about, I think earlier in the week we launched event filtering for all the Lambda based event sources.

Jeremy: Yeah.

Ajay: So that would be DynamoDB, Kinesis, SQS, and others. Essentially, you can now define filters on message attributes that allows you to control what events are being sent to Lambda. It's both sort of a throughput and cost optimization fee capability, allows you to get a lot more specific about the logic you want to push for.

Jeremy: Yeah.

Ajay: I think in the similar way we launched a partial batch support for SQS and Lambda together.

Jeremy: Yeah.

Ajay: Previously you had to kind of send the entire batch through to process events when you're building systems using SQS and Lambda together, you can do partial batches. Let's see, well, this ended up being a pre-announced but I'm sure it'll be something we get into the hands-

Jeremy: Yeah.

Ajay: ... of customers as soon as possible. We pushed one of the limits that customers definitely keep running into. We are going to be increasing the ephemeral storage space-

Jeremy: Yes.

Ajay: ... that we get as part of Lambda to 10 gigs.

Jeremy: 10 gigs.

Ajay: Previously, if you needed to push something like that, you would have to push your own EFS system, whole class of use cases where sharing EFS drives across different computer systems makes sense. But we also have customers who are like, "Hey, I just need ephemeral storage. I just need something that's built in with Lambda. Just push that number for me." So, we are now at the... We have 10 gigs, 10 GB RAM.

Jeremy: Yeah.

Ajay: 10 GB of memory with six virtual course.

Jeremy: Base 10, that's enough.

Ajay: Yeah. Base 10. Right. And then tens of thousands of concurrent executions-

Rebecca: Oh yeah.

Ajay: ... as you kind of get going, which is one. Although, I will say we have changed our pricing to one millisecond. So-

Jeremy: That's right.

Ajay: That's no longer on base 10, unfortunately. So we to break the pattern somewhere. I think all of the... I think that general collection's been really good from the core serverless perspective. I think I was really happy to see the broader announcements in Adam's keynotes with all the services adopting the serverless operational model.

Jeremy: Right.

Ajay: Redshift, MSK, EMR now all offer serverless flavors. So, it specifically means you spin down to zero or never pay for idle. You have automatic scaling built in based on usage, which for services like Redshift, it's super complex for us to be able to pull something like that off. And you're now getting that serverless operational model more and more pushed through all these services, which I personally am... This is one of those times where I'm really happy when the word serverless starts to mean a lot broader, because it's the right thing. The operation model is getting broader and broader over there.

Rebecca: So, that's a lot of launches.

Ajay: Yeah.

Rebecca: And week is not over.

Ajay: Mm-hmm (affirmative).

Rebecca: And also here you do more than like launch, right? You go to EVC, you talk to a lot of customers, you talk to potential customers, prospects like vendors, all sorts of people. I'm curious if there's... If you could tell us maybe about the best question you've heard so far, whether or not is the way the product functions, whether or not something about Lambda itself or one of the launches itself or from a customer in a specific use case, is there something that stands out to you where you're like, "Wow, we didn't think of that." Or, "We thought to that, but it's a complicated answer." How do you go into that?

Ajay: No, I hadn't thought of that. So thank you for thinking about it. It wasn't so much a question that came to me as a product owner, but a question I heard two technical leaders among the customers ask themselves. And what was fascinating for me was it's a question we have been trying to get customers to ask for a long time and now it just organically happens. Right? So, they were debating about the scale and availability benefits you got from a server full approach versus a serverless full approach. And they were talking about the various points of failures that you have to handle, and what happens when you run with Lambda, ECS too. And at one point during the question, one of the leaders looked to the other one and she goes, "Why wouldn't I built on serverless by default?"

Ajay: Right? And the other person said, "Well, that's a point to you. I don't have a good answer for it." And I'm like, "There you go. That's the question." That's the question that you all should be asked asking all the time is saying, "Why wouldn't I adopt a minimal footprint-

Jeremy: Yeah.

Ajay: ... service based approach to going in building and delegate as much as I can?" Because they were operating for speed. They were operating for agility, cost efficiency, elasticity as their business required at that point. And seeing that light bulb go off for them just based on the conversations we were having is very gratifying. Right? So launches all help turn nos to yeses or maybes into yeses. But hearing that organic like, "Why not?" Happen is really gratifying. So-

Rebecca: On a personal level, I just want to compliment you.

Ajay: Yeah.

Rebecca: Because I think you've seen that for a long time. Since 2014, you saw where... You saw that question being one of the end states. Like, "That's it. That's the question we want to get to, why wouldn't I?"

Ajay: Yeah.

Rebecca: But there's probably also a timeliness too, it took years to-

Ajay: Yeah.

Rebecca: ... building into that question.

Ajay: Yeah.

Rebecca: Do you feel like this is the first year that question is most resonant or are you like, "Oh, well it's actually been about three years." We've been asking that question?

Ajay: No, I would say that it's been a question we have prompted and customers have accepted and processed through. I would say through the COVID window is where we are seeing more and more customers go through this question by themselves because everyone's going through that realization of, "I have to be far more efficient, speed matters for me." Right? Like earlier when... The window when COVID started, we had so many companies including governments like the UK government, Indian government come out and say, "We're building serverless based systems for putting out COVID based testing and registration sites on you." That was the fastest way for them to build, handles the scale that they wanted to and run faster over there, right?

Rebecca: Mm-hmm (affirmative).

Ajay: And you see that kind of wheel click of saying, speed and cost efficiency has to come from this full approach is happening more. So I think the big shift is, it's gone from, "Why would I do servers?" To like, if we remind them like, "Have you thought about it?" They say like, "That's a good question." To them themselves asking the questions internally. And at that point that's the best part for all of us.

Jeremy: Yeah.

Ajay: Because if they're asking that themselves, then that's a discussion that doesn't need the serverless believers so to speak.

Rebecca: Yeah. Yeah.

Ajay: Right.

Jeremy: So, I've been watching the pre:Invent releases, some of the ones that have been announced, I'm hoping there'll be more tomorrow at Werner's keynote. But for me, I'm looking at all these releases and I think to myself, I'm like, "Okay, that's cool." Like the Kinesis one is cool and the Redshift one is really cool. And I want to talk a little bit more about that. And then it's like, "Well, now you can send S3 events to EventBridge." And those are... They seem like they're game changing in the sense that now I can... Even the event filtering, a great one where it's like, I can do things and I don't have to have as many moving parts. And I can get rid of some of that footprint like you said. But all of these launches to me seem very incremental. A couple more features.

Jeremy: Yeah. Maybe up the limits a little bit, a couple more things that can scale to zero. But for me, the transformative ones for serverless was obviously Lambda when that launched.

Ajay: Yeah.

Jeremy: But then Step Functions and EventBridge, those were like massively transformative things that changed the way that we wrote applications completely. And I'm just curious like, are we topping out? I mean, are we sort of like at the end of the serverless revolution here and now it's just we need to get people to use it or is there a next step?

Ajay: So, well, yeah. We have tapped into a debate which we've been having internally as well. So, I'm going to state something out there which we'll coin some phrases, right? So first look, I do believe when a product is going through the maturity cycle we're going with their... It matters more that we keep pushing out barriers for customers than anything else and making things more efficient, more scale.

Jeremy: Yeah.

Ajay: Like one of the benefits we've always talked about for serverless is it gets better with you doing nothing. Right? Now like, if you were running event filtering before and you were doing everything in code, you now just found a whole bunch of code to go delete.

Jeremy: Right.

Ajay: If you are previously paying for a hundred milliseconds, we dropped it to a millisecond, it's gone. You flip a bit on it, you deploy on ARM. The cost gets a lot better. This was one of the releases with Graviton and Lambda-

Jeremy: And Graviton, yeah.

Ajay: ... that earlier in the year, right? So, but I do think like if you go back to the original vision of serverless and Lambda, we had said, it's the fastest way for you to go from ideation to production, right? Minimal code.

Jeremy: Yeah. Yeah.

Ajay: Minimal operations, right? Like even the word... The famous, no ops frames-

Jeremy: Right. Right. Right.

Ajay: ... started out earlier on. And we move past that.

Jeremy: We do not support no ops here though.

Ajay: We do not support ops here, right? Exactly. That's a bad word. But I do think what's happened in the phase is we had that initial spot where we were like, "Hey, this is a great way for you to build." And I think what has happened is the last few years has been about making the operational model mainstream.

Jeremy: Right.

Ajay: And I think we are at that point.

Jeremy: Yeah.

Ajay: Like you look around no one questions, "Oh what is the serverless operation model?" If anything, you're seeing it being used in your stre... That term is being stretched in ways that we hadn't imagined when we started out. But the concept of abstract infrastructure scaling should be right. Or a spin down to zero should be right. Is just mainstream. And the fact that you're seeing serverless services like EMR and Redshift and Aurora and Kinesis, DynamoDB and every one of them kind of switched to that model.

Jeremy: Yeah.

Ajay: Is the momentum is there. Right? And I think that journey is going to continue, of like incremental capabilities to improve integrations. You're going to see more of these things. You brought up Lambda, Step Functions, and others as primitives that came up.

Jeremy: Yeah.

Ajay: And so, one thing we have talked about is, "Hey, is there a missing primitive over there?" Like we've talked about, I think you and I in a previous podcast, we talked about cash as a missing part.

Jeremy: Right. Yeah.

Ajay: Maybe something like that-

Jeremy: Yeah.

Ajay: ... will happen soon. But I do think, I'll borrow a paddle from the MCU over here, right? We are at the end of would what I'll call serverless phase one.

Jeremy: Okay.

Ajay: Right. We have established the primary players, the models understood. We have all accepted the premise.

Jeremy: Yes.

Ajay: Right?

Jeremy: Yeah.

Ajay: Just like in MCU, you accepted the premise like, "Yes, we have superheroes and they're part of our common lives." Suspension of this belief. We are all accepted that, "Hey, the serverless way works." Right?

Jeremy: Yeah.

Ajay: Like it's fast, it's cheap, it's secure, it works over there. I think we are at the beginning of phase two and I don't know what that looks like yet. Right. I mean, I wish I could predict, but I don't know what it looks like. And I think there it is. It's not about the operational model. It's about getting people to buy into the development model. Right?

Jeremy: Yeah. Yeah.

Ajay: We work so hard to bridge the development model they're used to, right? We're saying, "Hey, you like writing with code. You don't like Config as much-

Jeremy: Right.

Ajay: ... that's okay. Bring all the code you want-

Jeremy: Yeah.

Ajay: ... we'll make it easier." I think solutions like App Runner or Cloud Runner or others, that's the right thing.

Jeremy: Or even serverless cloud.

Ajay: Or even serverless cloud. There you go. Yeah. I love that service. I never thought of that one. They help you bring the way you build into the operation model that you're sharing. But then the ones who have also pushed the edge know that code becomes a liability over time, right?

Jeremy: Yes.

Ajay: Like the more you push that down to services for automation, the more you minimize your footprint of code, the more these agility and scale benefits come out. And I think serverless phase two or serverless V2 or serverless plus, right?

Jeremy: Serverless plus.

Ajay: Serverless plus is going to be about making that experience concept mainstream.

Jeremy: Yeah.

Ajay: Where people are used to taking service primitives, quickly assembling them to the app they want to. Community sharing of those patterns, bringing those patterns together and making that go faster. And what that's going to look like, I love all the innovation that's going on across in the community and ecosystem. They're like, serverless cloud is a great example. I like what the folks at Begin are doing.

Jeremy: Yeah, absolutely.

Ajay: The past Stackery and SAM and others have tried various flavors. I'm assuming that we could kind of play on those particular ones. And that kind of brings all of those pieces together for us to make sure that serverless we do is pushing into a flavor that we all can understand. Right? And I think this by the way is one of those things where I look at someone like Cloudflare.

Jeremy: Yeah.

Ajay: I like some the pieces they're doing, because in some ways they're one of the few players who are also trying to experiment with this serverless full development model-

Jeremy: Yes.

Ajay: ... and pushing into more customers and seeing what that looks like. Right? So, that's cool. I think as a community, we need more people trying to push that development model as opposed to only trying to bridge back. Right? It's one thing to bring people to where we are.

Jeremy: Right.

Ajay: But we also need to be building what's coming next.

Jeremy: Right. And I also think there was a lot of gravity with what AWS was doing where it's like, well, if I want a NoSQL database that's serverless, well, then I really have to choose DynamoDB because it was like the only one that was out there. And then you did some work and you created the... I'm blanking on the name, but the Cassandra version of it and DocumentDB and some of these other things. And then you had other companies that started following suit. So MongoDB has their Atlas, which is serverless. Astra DB, which is another Cassandra serverless option. And I think that that type of momentum now where developers can come in and say, "Well, I don't want to use DynamoDB. I want to use Cassandra." Or, "I want to use Mongo." Or whatever.

Jeremy: Now they can start using familiar tools with the serverless mindset behind it. And I think that's super important. The other thing I want to say too about infrastructure as code, because you say code is liability. Code is not an asset. The more code you write is the more code you have to look at. And this is one of the things that we're kind of what we're trying to do at serverless cloud, is we also think that infrastructure as code, when you have to write all of the mappings to the different primitives, that becomes a liability because all of a sudden now you go from DLQs to Lambda to destinations, and then all of a sudden you have to unwind a whole bunch of infrastructure's code and that doesn't update for you automatically.

Ajay: Yeah. Yeah. No, I love what you said about choice that customers need to pull out and making sure that it surfaces in an opinionated way that they have to build, right?

Jeremy: Yeah.

Ajay: I think the direction you have to end up is, or at least I would assert is you want the list of serverless primitives to keep growing.

Jeremy: Yeah.

Ajay: Like people need to have all the flavors of everything that they can make available over there, right?

Jeremy: Yeah.

Ajay: Like you have a slightly different model of programming model choices, use Fargate or Lambda. You have slightly different preference in data modeling, use DynamoDB or Aurora or Cassandra, whatever else that-

Jeremy: Right. Right. Yeah.

Ajay: ... touches your fiction, right? So it has to be huge selection of primitives, right? Including the tooling ecosystem for what it's worth, right?

Jeremy: Yeah. Yeah.

Ajay: Yeah like, you have all the vendors and people who are out there, right? I think you're going to have opinionated ways of putting them together though, right? Like I think that's where the experience comes in. And I think you're going to see more and more opinion come into experience. And when you're putting together services, that's where concepts like infrastructure's core come in, right?

Jeremy: Yeah.

Ajay: In fact, I've been kind of wondering whether infrastructure's code is a term we need to do away in the serverless world.

Jeremy: Yeah.

Ajay: Because we are almost invisibly teaching people that the services they're working with is actually infrastructure and it's not.

Jeremy: Yeah.

Ajay: It's application primitives. And that you're composing together as this thing. Right? So should we be calling it app composition, app integration.

Jeremy: Yeah. Yeah.

Ajay: Like you have talked about stuff like that in the past.

Jeremy: Yeah.

Ajay: And I do think that's a place where I hope to see some new stuff come up. Right?

Jeremy: Yeah.

Ajay: Like where infrastructure's code becomes more of the assembly language and there are higher level primitives, that sort of push on. We're trying something like that with SAM.

Jeremy: Yeah.

Ajay: I love what's going on in the serverless cloud space as well, but we'll see. It's phase two, we'll see how serverless plus, we'll see where it ends up.

Rebecca: So in a serverless plus world-

Ajay: Yeah.

Rebecca: ... do you see AWS producing more of that opinionation?

Jeremy: That's a good question.

Ajay: No, that is a good question.

Jeremy: Thank you.

Ajay: I don't think it'll ever be only AWS, right? I think AWS's success has always come from its partner and community and ecosystem. Like even the way serverless development has spread, it's required tools like Serverless Framework and others to say, "Hey, here's creative models for doing it." Right? Like I would be the last one to claim that AWS has a solution for everything. We are really good at creating the primitives and making them move faster. I do think we'll be sharing our own experience on how these primitives should come together. Like, we have large scale adoption within. There're other parts of Amazon. Like nearly half applications are built using Lambda in some form of fashion. So, you will start seeing more of that opinion show up in the way we develop, but not everyone wants to build like that. There will be teams who want to build like maybe like they want to build like Capital One, maybe they want to build like Steady.

Jeremy: Yeah. Right.

Ajay: Or they want to build like Brooks. They'll have different opinions of how that comes together. And that's okay. Right? Like I almost imagine that just like you have language selection today.

Jeremy: Yeah.

Ajay: Popular languages, you're going to have flavors of these serverless experiences and how the perimeters come together. Right? And just like nowadays we don't really... Like every language has their own way of expressing loops and variables and states and everything, but ultimately we all understand what loops and states and variables look like. Just like that. I imagine a world where functions, workflows, queues, and EventBuses will be just well understood primitives that you can kind of treat differently, but the way you compose them will have idiomatic stuff that works with your language and otherwise.

Jeremy: Right.

Ajay: And so, no. I do not see AWS as the only one doing it. I hope that our primitives perform the best in terms of scale availability, latency, and capacity. And our experience is one that helps customers move faster, but won't be the only one. So-

Jeremy: So, speaking about Steady. So Zach Canter, I think had a tweet at one point that said the next AWS would be built on top of AWS. And I think that's an interesting thing because the question is what is AWS's job? What's their role in this larger ecosystem? As you said, or as the question that Rebecca asked is, if you're focused on these primitives and making these primitives compatible with all of these potentially opinionated things, because you've got enterprises that want this, and you've got another company that wants that and you've got this really broad ecosystem where everything has to integrate. I'm just really curious where AWS falls into that because right now, if you said to me, "Hey, I need to process a webhook from anywhere." And I need that to do some processing and put it into a database.

Jeremy: I think I can come up with about seven or eight ways that I can do that right now, all with serverless capabilities. And the question is going to be what's the right one to choose and when's that the right one to choose? And again, I think it's part of the organizational structure of AWS with the two pizza teens and a lot of people working, not working as cross-functionally maybe as they could be. But again, with an organization this size I think it's hard to do. But I'm just curious like, that type of... Or maybe this isn't a question, more of a wish a wishlist. More of a wishlist is like, I think as more of these opinionated frameworks come out, what Sean Wang calls a self provisioning run time, that these sort of things... That AWS will work with these frameworks to say, "Well, what that you need us to do so that you can make the right connections and you can speed up that developer experience?"

Ajay: Yes.

Jeremy: I know. Happy birthday.

Ajay: So, I-

Jeremy: Thank you everyone.

Ajay: Yeah, I already know.

Rebecca: What?

Ajay: No. Look, so to two call outs over there, right? And I think you've heard Adam speak about this in his keynote as well.

Jeremy: He did.

Ajay: Right? AWS has to do more to help build both horizontal and vertical experiences-

Jeremy: Right. Right. Yeah.

Rebecca: Yes.

Ajay: ... to customers. Right. And I think you're going to see the same flavor when you're talking about developer audience and administrative audience and otherwise, right?

Jeremy: Yeah.

Ajay: So, I think you're going to see more of these opinionated approaches come together with narrower focuses, right?

Jeremy: Yeah.

Ajay: One of the asks I heard out here was saying, "I wish AWS had a personalized button that only showed me the services that I use."

Jeremy: Right.

Ajay: And I'm like, I've never thought of that. Like, yeah.

Jeremy: Don't I personalize into the search bar though.

Ajay: No yet.

Jeremy: Because you'll get a different service that doesn't do what you think it might do.

Ajay: Right. Well-

Jeremy: Skills of naming aside. I'm sure that some flavor of personalization on AWS would be something that all of us would enjoy seeing. And I think that kind of view is something I would hope that as part of the simplification of selection, as part of building that opinion, you will start seeing big more and more. Right? Yeah.

Ajay: You see some of this with customers doing this with like things like CDK Patterns.

Jeremy: Mm-hmm (affirmative).

Ajay: We have serverlessland.com where you're saying, "You know what? If you know nothing and you're starting with these patterns, here's the simplest way for you." And doing. So, we are in the early days. Like I said, serverless plus is just getting started.

Jeremy: Right. Right.

Ajay: So right. So a year from now, hopefully we are back here and we're like, "Hey, we talked about these things. Look-

Jeremy: Right.

Ajay: ... we're all getting to grow up a little bit."

Rebecca: Yeah.

Jeremy: But I think you're right though that baseline or the sort of the infrastructure structure that's needed, the operational model is there.

Ajay: It's there.

Jeremy: And now it's just on all of these other people out there to start figuring out, what are the patterns to do all this stuff in a way that you can deploy these patterns and implement these patterns without going through this potential pain point of writing all this configuration?

Ajay: I agree. So, I will say as a provider of these services, and this is not just AWS.

Jeremy: Yeah.

Ajay: I would put this out as a call to action for anyone over there.

Jeremy: Yeah.

Ajay: You have to make sure our primitives need the promise that we are all putting there.

Jeremy: Yeah.

Ajay: People are trusting us with their security, with their scale-

Jeremy: Yeah.

Ajay: ... with their latency, and others. And we have to keep making that better. We have to provide governance hooks. The hooks for observability tools to plug in the books-

Jeremy: Yeah.

Ajay: ... for all these things. And it's not so much that those have to be standardized APIs, but there need to be APIs. Those need to be interfaceable. Those need to be customizable.

Jeremy: Yeah.

Ajay: So that as people build opinions, they can build the opinions they want to bring and bring that forward. I feel like the wrong path to go down here would be you end up building these services that only work together a certain way.

Jeremy: Yeah.Yeah.

Ajay: Or only work with the certain set of services. One of the things we have worked quite a bit on over the last few years is loosening some of those opinions, things like extensions.

Jeremy: Yeah.

Ajay: Even the fact that we are doing contain an invincible. Right?

Jeremy: Yeah. Yeah.

Ajay: A lot of that was like helping people come from where they are. But I do think those rules carry forward even when you're looking at what people need to build next.

Jeremy: On that.

Ajay: Yeah.

Jeremy: Do you have another question? I have one more question for when we're running out of time. You're our first guest.

Rebecca: I do. I have final question.

Jeremy: Okay.

Rebecca: I think mine the final though, so it's like you're going first. Yeah.

Jeremy: Yours is the final, okay. Mine's almost the final question too. So Kinesis.

Ajay: Finals, yeah.

Rebecca: Two finals.

Jeremy: Kinesis. They announced on demand Kenisis at the keynote yesterday. And I saw it. I saw that it was now price I think per second, I think for the streams. And so I'm looking at this and I'm like, "All right, finally, I can tell Chris Munns that Kinesis is serverless." And then I think, Ben Kehoe corrected me.

Ajay: Yeah.

Jeremy: "We're not quite there yet." Huh?

Ajay: Quite there yet. But it's always good to go and correct Chris Munns, that's always a fun site to see. I have certainly had my experiences on both sides on that. Look, I think there's value in what Kinesis launched even as non-

Jeremy: Yeah.

Ajay: ... non capability, right? Like the auto scaling capability of the Granular Pricing is something that gets over there. Kinesis is a complex and rich footprint for us to go and do so.

Jeremy: Yeah.

Ajay: And I'm confident that with the engineers who are working on it, they're aware of the feedback customers are giving and eventually you will get... My hope at least is you will see some flavor of spin down to zero.

Jeremy: Yeah.

Ajay: And like if you had asked me three years ago, would Redshift end up with a serverless offering-

Jeremy: Yeah.

Ajay: ... and I'm like, "I have no idea how you would even do that."

Jeremy: Yeah.

Ajay: And yet here we are.

Jeremy: Yeah.

Ajay: So never say ever.

Jeremy: Right.

Ajay: But at least, let's make the most of on-demand Kinesis that we have.

Jeremy: Yeah.

Ajay: And fingers crossed in a full floor.

Jeremy: And I got to play around a little bit-

Ajay: Yeah.

Jeremy: ... but I feel like there might be some automation you could put in there to turn those streams off when you're spinning down the development environment or things like that. So-

Ajay: so-

Jeremy: ... you might be able to work it in. But yeah.

Ajay: So, I think you should think about more of this as it's spin down to one-

Jeremy: Yeah, right.

Ajay: ... rather than down to zero and decadence.

Jeremy: Good point. Yeah.

Ajay: And so-

Rebecca: Yeah.

Ajay: ... just like with other systems and when you talk about spin down to one-

Rebecca: Yeah.

Ajay: ... there are benefits. It's always running. You always get started and instantaneously scale benefits, optimizations that way.

Jeremy: Yeah.

Ajay: The downside is that the cost number hangs out for a little bit longer than you would like to. But if there's one thing AWS is good at, it's improving cost structures as we go along. As we have seen with-

Jeremy: Right.

Ajay: ... even some of our recent announcements.

Jeremy: That's a good point. Yeah.

Ajay: Yeah. Right. Rebecca?

Rebecca: So, as the leader of Lambda-

Ajay: Yeah.

Rebecca: ... Experiences, maybe you could say you're the father of all these launches, are you allowed to have a favorite? And if so, what's your favorite?

Ajay: Ooh. So, am I allowed to have a favorite? I get giddy about most of them. Right? Like I remember when... So, I'll maybe say two, if that's okay. Right?

Rebecca: Okay.

Ajay: So one, I really like... This is actually not a Lambda launch, we don't actually talk about it. So, SageMaker launched a serverless inference that's powered by Lambda. And I love the broader story of Lambda being an invisible component-

Jeremy: Yeah.

Ajay: ... that just makes things work and brings the serverless model elsewhere.

Rebecca: Totally.

Ajay: We have talked about powered by Lambda in other cases, we've had external systems that ratify functions and others. SageMaker doing serverless inference on us. And ML inference is such a populous scenario right now.

Jeremy: Yeah.

Ajay: Like having that as smooth scenario, love the slickness that's coming. There's more, it's just in preview. My other one was actually SAM Accelerate. Right? Because it's right there like this-

Jeremy: Right.

Ajay: It's part of that serverless plus story-

Jeremy: Yeah.

Ajay: ... where you go and say, "Yes, I'm opinionated about the fact that local emulation is not something you do as much. You test within the cloud."

Jeremy: Yeah.

Ajay: But it's done in a way that brings sort of that old experience back to the new. It gives people a path to the future.

Jeremy: Right.

Ajay: As opposed to just saying, "I'm going to meet back to where you are." So I love what the team launched over there. Every time I see the demo and I played with it, it was just like, "Yes." This is like, "I love this similar experience that you're doing with serverless cloud." So yeah, that those are the two that would jump out. But-

Rebecca: You handled that so gracefully. That was such a like, "Are you allowed to pick?" You're like, "Well, I got two."

Ajay: Maybe you must have PR training.

Rebecca: You thought about this for a long time.

Jeremy: So, I just-

Ajay: I just love this.

Jeremy: Before I let you go, I will say my, I think my favorite launch so far was the serverless Redshift.

Ajay: Yeah.

Jeremy: Because I really, really love Redshift as a data warehouse. We've used it in a number of companies I've worked with. And then when I went to a startup, the cost was just prohibitive to really store all the data that we wanted to store there and have those clusters constantly running. And I love the fact that now you can get row level update, which you can't necessarily do with S3 to Athena. So I think there's some interesting patterns there as well. So yeah. That serverless pattern's site is going to have to add quite a few-

Ajay: Quite a few.

Jeremy: ... quite a few new ones.

Ajay: No, that's what the new year is for.

Jeremy: Perfect. Anyway. All right.

Ajay: Well.

Jeremy: Ajay, thank you so much for being here.

Ajay: I'm honored.

Rebecca: Thank you so much.

Jeremy: Enjoy the rest of re:Invent. I'm sure there are probably some more launches tomorrow at-

Ajay: Yeah.

Jeremy: ... Werner's keynote that we will find out about-

Ajay: Yeah.

Jeremy: ... altogether. So-

Ajay: Like I said, watch for the post:Invent season. They'll be more coming there too.

Jeremy: Perfect. All Right. Okay. So everyone watching, we're going to take a quick break and then we're going to be back with Talia Nacy, we're going to talk more about SAM Accelerate and a bunch of other really cool things.

Ajay: Okay.

Rebecca: And we're back with Serverless Chats, with our next guest Talia Nacy.

Talia: Hello.

Rebecca: Say yes.

Talia: Hello. Of course, thanks for having me. I've heard about this podcast from so many people and I feel honored to be a part of it, certainly.

Jeremy: We are honored to have you here.

Talia: Thank you.

Rebecca: Yeah, absolutely. So first of all, it's a very special moment and time in Vegas because your dad's birthday.

Talia: It is.

Rebecca: Yes, it is.

Talia: Happy birthday dad, if you're watching.

Rebecca: Happy birthday, Talia's dad. Yeah.

Jeremy: So maybe we can start, just give us a quick overview. Tell us what you do at AWS, what you do with serverless.

Talia: Yeah. So I'm a senior developer advocate with AWS serverless. And I started on the serverless team in April. So I'm relatively new to this space, but I got started with developer advocacy a few years ago. And so at AWS, what I do is I basically teach developers how to become better with serverless. So I speak at conferences and I write blog posts and I advocate for developers and how to make their experience better with serverless technologies. And this is my first re:Invent. So I'm-

Jeremy: Oh, wow.

Rebecca: Welcome.

Talia: This is a momentous occasion.

Jeremy: Are you finding it overwhelming yet?

Talia: Yeah. I haven't really slept. I kind of keep forgetting to eat.

Jeremy: That's right. I did not eat lunch today either. I believe so.

Talia: Yeah.

Rebecca: Before we dive into too much of the launches and the things that you've been talking about. So you were like a QA engineer.

Talia: Yes.

Rebecca: And engineered in a lot of different places before. That sounds bad. You've been an engineer perhaps before you became a developer advocate.

Talia: Yes.

Rebecca: And I think there's almost a leap that you might have to take at least from my what hear from others. Right? What it's like to be on one side and then the other.

Talia: Yeah.

Rebecca: And can you talk a little bit about that transition, right? Where you're like an engineer whose maybe looking to developer advocates and now you're on the other side.

Talia: Yeah.

Rebecca: Whether or not you're like, "I'm in." Or like, you're like, "Oh, this is hard."

Talia: Yeah. So yeah, I was a test engineer for a few years and as I was doing test engineering, I was learning more and more about what made testing effective for me. And one of the things that I felt really passionate about was testing in production. And that I kind of advocated for that at one of the companies that I worked for. And it ended up that like I would speak in tech talks and do women in tech meetups and stuff. And the traction just kind of started rolling. And then I submitted that talk to a conference and I kind of just became a little... I got a little momentum with public speaking because of that talk. And so I got a lot of good feedback from my public speaking and then, the company that I worked for after I started doing test engineering, they were like, "Oh, we want you to become a dev advocate because you're already doing it." Yeah. And then I switched over to Amazon.

Rebecca: Amazing.

Talia: Yeah.

Jeremy: So speaking about testing and production-

Talia: Yeah.

Jeremy: ... maybe not always testing in production, but certainly testing in the cloud.

Talia: Yes.

Jeremy: This has become a topic that is... That I'm passionate about, something we're working on. And it was funny because as we're basically about to launch what we are doing, I get a call from Farra, connects me with Stefano and we start talking about SAM Accelerate.

Talia: Yes.

Jeremy: Before SAM Accelerate launched. And basically this idea that you could write code locally, you hit save and as fast as possible that code gets into a test account in the cloud that then you can actually connect to all these different services. So tell us a little bit about at SAM Accelerate and maybe as a former test engineer-

Talia: Yeah.

Jeremy: ... why you're so excited about being able to test in the cloud.

Talia: Yeah. So, the whole idea of testing in the cloud is kind of the same idea as like testing and production. Like I don't care if my software works when I'm emulating the cloud because it's not the cloud, it's the same way like I don't care if my software works in a staging or test environment. I care if it works in the actual place where my software lives.

Jeremy: Right.

Talia: So with SAM Accelerate you're actually testing against the cloud, and it's super, super, super fast. I think there's like a 100 millisecond response time. So as soon as you save your changes, sam sync will watch for those changes and then you can test your changes almost immediately-

Jeremy: Yeah.

Talia: ... in the cloud. It's just super, super fast. And so there's sam build, which builds your code. And then-

Jeremy: Yeah.

Talia: ... sam sync, which syncs your code to the cloud. And then when you want sam sync --watch, you can just have that running and every time you save a change, it'll just push that up. And it makes testing just so much easier.

Jeremy: Nice. Now, have you found... Now, what is the reception to SAM Accelerate? How are people feeling about this new... Like I don't need to install LocalStack anywhere. And again, nothing wrong with LocalStack.

Talia: Yeah.

Jeremy: And I keep saying this over and over again, I think Brian Loru is watching us right now. And I want to say, I absolutely love Begin and I love Architect and I love what they do to emulate that local environment for you, because it gives that developer that soul package. And I think that's an awesome step for a lot of people that need that. But if there's a way for us to get things into the cloud so that we can test against on demand Kinesis and we can test against serverless Redshift and we can test against our SQS ques, and all those other things can happen. I think that's where we want to go. Right?

Talia: Right. Right. And to your point, I think when we make it as close to production as possible we're really telling the developers how... Giving feedback of how your feature is actually going to work. And so the feedback that we've gotten I... First of all, I don't think a lot of people know about SAM Accelerator yet because it's such a new-

Jeremy: Right.

Talia: ... feature that's released. I think we released it like two weeks ago, three weeks ago, maybe. So I don't know that a lot of people know about it. I have heard great things on social media. People are responding really well to it, but I hope that after people listen to this that they'll go and try SAM Accelerate. I don't know that we've gotten like a ton of feedback yet, but I've heard good things on Twitter. So-

Jeremy: Very cool.

Talia: Yeah.

Rebecca: How is the development experience for you? Were you involved... As a developer advocate, were you also involved in the product conception and growth of SAM Accelerate?

Talia: So I think SAM Accelerate has had the... The birth of it. Yeah. Conception of it started a long, long time.

Rebecca: Couple years ago.

Talia: Yeah.

Jeremy: Yeah.

Rebecca: Yeah. Yeah.

Talia: Way before I was at AWS. But I know on my team, Eric Johnson was very, very tightly, tightly close to this. So, hi Eric, if you're watching. Yeah. This is in Eric's wheelhouse, I know he was a part of that.

Rebecca: Yeah, because I'm wondering if they got to test on you all-

Talia: I'm sure-

Rebecca: ... before.

Talia: I'm sure they did. And also AWS, wouldn't go through all of that if it wasn't something that customers were asking for.

Rebecca: Yeah, absolutely.

Talia: So, it's definitely something that we had to get feedback for that like customers asked for it and they wanted it and that's why it was one of the biggest things that we launched.

Rebecca: So you recently, just yesterday actually, I think spoke on stage with David Richardson.

Talia: Yes.

Rebecca: Will you tell us a little bit about what your specific role was-

Talia: Yeah.

Rebecca: ... on stage?

Talia: Yeah. So, okay. So there were two demos that I did for the serverless keynote. The first demo I did was SAM Accelerate, where I demoed a few different commands. I demoed sam build, sam sync. And then I did sam sync with state machines. And then the second demo I did was Step Functions Workflow Studio, the new visual workflow studio which, I think in every single talk that I've done at re:Invent, five total, all of them I talked about the Step Functions Workflow Studio. So it's a really impactful feature, I think. For those of you who don't know the Step Functions Workflow Studio is a new visual experience that gives developers the ability to visually and intuitively build your workflows. And it's just an easy intuitive UI to understand how to connect different services. So you don't have to worry about like configuration code or workflow definitions. You just focus on your business logic and you don't have to worry about like how to connect the dots.

Rebecca: So I know that you... Jeremy, I'm cutting you off. I have one more.

Jeremy: No. Okay. No, go ahead.

Rebecca: So I know that you prepared quite a bit for your presentation yesterday on stage.

Jeremy: Yeah.

Rebecca: And I do think that there are plenty of engineers who would also be wonderful developer advocates.

Jeremy: Yeah.

Rebecca: But there's like something that's a little scary, and I think public speaking for a lot of people is quite scary. So maybe do you have any tips about like, "Hey, when you're about to get on stage for the first time." I know that wasn't your first time.

Talia: Yeah.

Rebecca: .. but, or if you're about to get on a big stage. Is there anything that you do to help you feel like really confident to get on that stage?

Talia: Yeah. So there's two things. One you would just have to be really, really prepared. The more prepared you are, the more you rehearse, the more you practice. That's going to be a good thing. And also I feel like you should only really speak about things that you're really passionate about. If you're passionate about what you're talking about, you're not going to be as nervous as if about something that you don't really care about. So I feel like the more that you are into it, the more that you're passionate about what you're talking about, it's going to make it a lot less scary.

Jeremy: So I was going to say as proof of life or the proof that we're live, Brian did get back to me. He says, "Thanks, Jeremy. Yeah, we totally agree. These things are complimentary. SAM Accelerate is brilliant. And we are very much looking forward to deleting our emulation layer someday."

Talia: Wow.

Jeremy: So that is-

Talia: You know.

Jeremy: But I think that is the future. And thank you Brian for that. Because I think that is the future. That is the idea to say, "If you can quickly get code and infrastructure to spin up and be in that environment for you, the fidelity there is just amazing." Right. But again, I want to say I love what LocalStack did because they filled the gap. It's so hard for a lot of people who are building surplus applications to get their code into a state where they can actually test it.

Jeremy: And it's such a great forward step to that. The other thing I wanted to say, because you mentioned the Step Functions Workflow Studio. Okay. I'll get it. But yeah. So the idea behind that and just this overall user experience, and I know you don't work on the Amplify team. Ally and a bunch of other people, we had them on the show to talk about the amazing things that Amplify is doing. But really for you, are you seeing that developer experience? These pieces are just starting to fall into place. We were talking a little bit with Ajay about this.

Talia: Yeah.

Jeremy: This idea of like, where's AWS's role? But I mean, what are your thoughts on where this is going? And you feel like this is all going in the right direction?

Talia: Yeah, absolutely. So what happened was in my breakout yesterday, there were two themes that I kept coming back to. Serverless development is getting easier and faster.

Jeremy: Mm-hmm (affirmative).

Talia: With every launch that the serverless team releases, I feel like it does one of two things. It either makes serverless development easier or faster.

Jeremy: Yeah.

Talia: And I talked about that a lot yesterday in my breakout, and the Step Functions Workflow Studio is a great example of that where it's going to speed up how you're developing your state machines. And it makes it easier because it has like this great visual UI. Just another way that serverless development is becoming easier and faster.

Jeremy: Right.

Talia: And same with SAM Accelerate. I mean, it's so much easier to test. It makes it faster to test. I think those are the two themes that we're going to continue to see in serverless development.

Jeremy: I love that.

Talia: Yeah.

Rebecca: Since you've been... This is your first raiment, so I'm sure you're... I mean, I'm always overwhelmed no matter how many years I've been here. Is there any emerging trend that you're hearing from either customers or vendors or at breakout sessions or workshops that you're like, "Oh, I think that's where we need to go next." Have you kept hearing kind of the same pain point or the same drum beat, I guess.

Talia: So what I've heard from my colleagues is that, previously at re:Invent a lot of the questions that people have gotten have been more like low level like "What's serverless?" Like, "How do you develop with Lambda?" Like very kind of broad questions. And the questions that we're getting now are very, very deep, like 300, 400 level, like, "How do I build this thing with this service and this?" And they're very much more complex now. I feel like that means that people are like utilizing serverless more and more and they have like more hands on. I mean, I feel like the past couple years with COVID everyone just is in a deep learning.

Jeremy: Right. A lot of videos, a lot of content.

Talia: Exactly.

Rebecca: Right.

Jeremy: Yes, absolutely.

Talia: So, I think people are getting into it a little bit more are now, and it's not about like, what's serverless? Or what's Lambda? It's more about like, "I want to build this specific thing with these services and tell me what the best way is." And-

Jeremy: All right. So while speaking of some deep 400 level questions-

Talia: Yeah.

Jeremy: ... let me see if I can throw one at you definitely.

Talia: Yeah. It was so deep. Yeah.

Jeremy: Event filtering-

Talia: Yes.

Jeremy: ... which Ajay mentioned briefly, tell us what event filtering is.

Talia: Yeah. Okay. So, with events that get put onto a stream, so I'm talking about Kinesis, I'm talking about DynamoDB streams, an SQS stream. So basically anytime before event filtering, anytime an event was put onto a stream and put into the Lambda function, every single event that was put onto the stream would trigger the Lambda function. Now with event filtering, you can say, "I only want this type of event to trigger that Lambda function." It could be like an attribute of the event. It could be like something inside the event. It could be like the... The example that I gave yesterday was, think of like a peloton race, in a peloton race you have events firing at every second and each event has a ranking of how you're doing in the race. So like at second five I could be ranked 12, but at second 10 I could be ranked three. So-

Rebecca: One.

Talia: Right. Yeah. And so let's say that I wanted to do something where I wanted to get the ranking at the end of the race.

Jeremy: Yeah.

Talia: And I didn't care about the ranking in the beginning. So before event filtering every event that would come onto the stream would trigger the Lambda function and you're paying for all of those implications.

Jeremy: Yeah.

Talia: So now with event filtering, you can filter so that if you have like a final attribute, if you're only checking for final equals true, then all the other events don't matter.

Jeremy: Right.

Talia: Lambda doesn't care about like this second, this event, it only cares about the final event. So it's going to dramatically reduce the number of times your Lambda function is invoked. It's going to save you money. It's going to cause your whole applications to run faster-

Jeremy: Right.

Talia: ... overall.

Jeremy: So-

Rebecca: Is there any like paradigm shift you have to educate people about when starting to use event filters?

Talia: I don't think it's necessarily with event filters. I think it's just events in general are so different than like APIs. So it's just a different way of thinking. I think once you nail events and you really understand the concept of an event, then event filtering fits into your workloads more than you would if you were using like APIs, for example.

Rebecca: Yeah.

Talia: But I think it fits really nicely into event driven architecture.

Jeremy: Yeah. So I'm going to get into the 400 level question now.

Talia: Exactly. I think you have another 400.

Jeremy: Okay.

Rebecca: Same here. Yeah.

Jeremy: Well, so now this... There was more workarounds in the past. So before you'd be reading off a DynamoDB stream, I would read every event into that DynamoDB stream. And then if I wanted something else to happen as a result of a created or it matches some event pattern whatever, I would go ahead and I would maybe use the Lambda API, and I would do an asynchronous invoke somewhere else.

Talia: Yes.

Jeremy: Or put it on EventBridge.

Talia: Or you do it in the Lambda function. A lot of people have done that.

Jeremy: In the Lambda function I'd write a whole big-

Talia: Right.

Jeremy: ... big thing and I wouldn't be able to isolate it. And maybe I get a lot more creates than I get updates and deletes and-

Talia: Right.

Jeremy: ... wow I have to scale all those things, I'm competing with services. Maybe I dump into EventBridge, and then I lose ordering and all kinds of stuff like that. So the interesting thing about the way that this works now is I think, maybe this isn't a question, maybe I'm just making a statement.

Rebecca: This is Jeremy's favorite.

Jeremy: This is what I tend to do. I'm like, let me see if I can find a question in there. No, but so this is something that I figured out too is, when you say you save money because you're only invoking it when those events come down, you're also saving money and that you're not reading everything off the stream anymore.

Talia: Right.

Jeremy: And I've confirmed this with a few people that you're not everything off the stream anymore, that the Lambda folder itself will eat all those costs and only charge you for the ones that actually get delivered.

Talia: Right.

Jeremy: So you could be using Kinesis stream for example, for a lot of like clickstream data, but you're only interested in certain events that you want to do with Lambda. Now you may have five or six Lambdas each one does something different, all reading off the same stream, but you're only getting charged for the actual ones that invoke the Lambda functions-

Talia: Right.

Jeremy: ... as opposed to everything. So-

Talia: Right.

Jeremy: And also, I haven't played around with batching yet. Oh, there's my question. How does batching work with the event filtering?

Talia: Yeah. So right now, what happens is you filter your events and then your events get batched up into one-

Jeremy: Oh okay.

Talia: ... into where they're supposed to go. And then it gets sent to the Lambda function.

Jeremy: Ah, so you can still... So if I still wanted like a batch of a hundred or whatever,-

Talia: Yeah.

Jeremy: ... still it'll compile those, even this.

Talia: Yeah.

Jeremy: Okay.

Talia: Yeah.

Jeremy: Very cool.

Talia: All right.

Rebecca: Can you also... This might be a very 100 question, so sorry if I'm taking it back there but-

Jeremy: No, that's fine.

Rebecca: Can you also side to say event filtering on this peloton race, I want to be able to return this every 10 seconds or every one minute-

Talia: Yeah. Yeah. Yeah.

Rebecca: ... but I don't want to return it every second.

Talia: Yeah.

Rebecca: And then to the end user do they end up only... They just feel it and saying like, "Oh, it takes 10 seconds to update before it tells me."

Talia: Yeah. You can configure that.

Rebecca: All that.

Talia: Yeah.

Jeremy: In batch sizes. In like batch time windows or something like that. Is that there-

Talia: Yeah. I think there's a default and you-

Jeremy: Right.

Talia: ... but you can configure it.

Jeremy: Yeah.

Rebecca: That makes sense.

Jeremy: Interesting.

Rebecca: Everything configurable.

Talia: Yeah.

Jeremy: All right. So-

Rebecca: I think you had another 400 in there, Jeremy.

Jeremy: I maybe had another 400.

Rebecca: I want to keep up that last one for you.

Talia: Yeah, the hardest question.

Rebecca: Yeah.

Jeremy: What did I want to talk about? Oh, so we didn't get a chance to talk about all the different with Ajay because we haven't kept it a little bit higher level, but what about like using batch publish to SNS topics that we've finished.

Rebecca: We did actually get to where we were going. Okay, good.

Jeremy: Okay. All right.

Talia: I know with batch publish now there's a new batch publish API for SNS-

Jeremy: Yeah.

Talia: ... that matches what SQS did a while ago. So now you can publish messages batch with SNS and again, it saves you money and it makes your development faster.

Jeremy: Right. Right.

Talia: So there's two days left. No, one day. What day is it today? Wednesday.

Jeremy: Today is-

Rebecca: I don't even know.

Talia: I know there's like two days left. I don't know. Today is-

Rebecca: Where are my people find you for the rest of re:Invent.

Talia: The rest of... Okay. So I'm doing a workshop tonight and the workshop is a-

Rebecca: It's a small one, right? The hands on workshops where it's like 50 people.

Talia: No, it's like a 400 person workshop.

Rebecca: Oh wow.

Jeremy: Wow.

Rebecca: The party. You're hosting a party tonight with computers.

Talia: Yeah. Okay. It's on serverless image processing.

Jeremy: Oh.

Rebecca: Nice.

Jeremy: Very cool.

Rebecca: We're using the new visual workflow studio from subfunction.

Jeremy: Very nice.

Talia: I know where I'm supposed to be at what time, but-

Jeremy: So what time is that and where is it?

Talia: It's 6:15 at the Venetian on the second floor.

Rebecca: Okay, cool.

Talia: Yeah.

Jeremy: Love it. Well, all right. So we're running out of time. So we just have to tell our listeners, well, first of all, thank you so much-

Rebecca: Yes, thank you.

Jeremy: ... for being here.

Talia: Yeah.

Rebecca: Thank you so much. And welcome to AWS serveerless-

Talia: Yeah, exactly.

Rebecca: ... developer, yeah.

Talia: It's always great. I love that. Getting more people on the serverless tea and people, especially people pushing testing in the cloud. I love it. I love it. I love it.

Rebecca: Yes.

Talia: Yeah.

Jeremy: Anyways, well, we have an announcement.

Rebecca: Good training majors, testing and production.

Jeremy: Right. Yeah. Testing and production.

Rebecca: Deploy, and Friday testing and production.

Talia: Yes.

Rebecca: So Rebecca.

Talia: Deploy at any time.

Jeremy: Right.

Rebecca: Yeah.

Jeremy: So Rebecca and I, we are going to take a break for the winter. We're going to take to holidays and all that kind of stuff.

Rebecca: We're going to a bear den.

Jeremy: Yeah. Right.

Rebecca: We're going to eat a lot of stuff in the hibernate.

Jeremy: Right. Well, I don't need to eat anymore. COVID and bread killed me. So, but well, we are going to take a break for two months and we're back in February. And we're going to have all new episodes and I don't know, cross your fingers for Werner, we've been trying really hard.

Ajay: Add Werner. If you are watching.

Jeremy: It-

Ajay: Travis, if you're watching.

Jeremy: Travis and Lisa.

Rebecca: Lisa.

Jeremy: Right. Good. Exactly. So it could happen. We're super excited about it.

Rebecca: Well, a few things.

Jeremy: Yeah.

Rebecca: One, thank you to all of our guests.

Jeremy: Absolutely.

Rebecca: Thank you to all of our listeners. Thank you Jeremy, for asking me to be a co host.

Jeremy: Thank you for accepting.

Rebecca: And of course, yes. And then when we come back in February, if you all have guests that you would love to hear from.

Jeremy: Yes.

Rebecca: Werner. Please let us know because we are preparing and we'll be prepared in February.

Jeremy: Yeah.

Rebecca: But we are happy to hear from you all who you want to hear from.

Jeremy: Right. Amazing.

Rebecca: Anything else you want to add? I totally just-

Jeremy: No, just, I mean, it's been great knowing you. We've done, what? 15 episodes this fall.

Rebecca: 15.

Jeremy: Something like that.

Rebecca: Something like that. 20, 18.

Jeremy: It's 28, 100. I don't know. It's been a lot, whatever it is. It's amazing. But this has been super fun. I'm looking forward to keep doing this with you and having more great guests like Talia and Ajay and just Werner and all kinds of things like that. And it's-

Rebecca: Say for Werner for us.

Jeremy: Yeah, it's-

Rebecca: It's Werner. Add Werner.

Jeremy: It's going to be-

Rebecca: Paging. Paging.

Jeremy: Great. It's going to be great. So anyways, thank you everyone. And we'll see on Twitter, but we'll be back with new episodes in February.