Sign up to save your podcasts
Or
Share Episode #65: Serverless Transformation at AWS with Holly Mesrobian
Share to email
Share to Facebook
Share to X
Share to email
Share to Facebook
Share to X
Or
Episode #65: Serverless Transformation at AWS with Holly Mesrobian
About Holly Mesrobian
Holly Mesrobian is a Board Member at Cascade Public and the Director of Engineering for AWS Lambda. Holly has 25 years of experience in designing, building, and managing globally distributed teams in software development, and more than 15 years as a leader of leaders. She has in-depth experience with building services for builders, and for wireless and broadband carriers; online services for direct to consumer offerings; and commercial shrink-wrapped software. With a double Master’s Degree in Computer and Science and Software Engineering, Holly began her career as a developer before holding leadership positions at companies, like Amazon and RealNetworks, and startup Cozi.
- LinkedIn: https://www.linkedin.com/in/holly-mesrobian-a1b710/
- Under the Hood of AWS Lambda 2019: https://www.youtube.com/watch?v=xmacMfbrG28
- Under the Hood of AWS Lambda 2018: https://www.youtube.com/watch?v=QdzV04T_kec
Watch this episode on YouTube: https://youtu.be/nBYUh7CVUiQ
Transcript
Jeremy: Hi everyone. I'm Jeremy Daly and this is Serverless Chats. Today, I'm chatting with Holly Mesorbian. Hey Holly. Thanks for joining me.
Holly: Hi, thank you for inviting me.
Jeremy: You are the director of engineering for AWS Lambda at Amazon Web Services. Why don't you tell the listeners a bit about your background and what the director of engineering for AWS Lambda does?
Holly: Absolutely. Engineering leaders in Amazon are very technical and I think I fit in that class of leader. I've been in engineering for more than 25 years. The first decade that I was in, I was actually an engineer, and then the last 15 years or so, I've been leading large-scale engineering organizations that are also responsible for 24/7 operations. You think about those, they're called DevOps organizations. That's what I've been doing for quite a while now. The Lambda engineering organization is just like that. In terms of my background and how did I get here?
I have two graduate degrees, computer science and software engineering, and as I referenced lots of time, designing and building systems. One of the things that's really great about AWS and leading the teams here, I reference that DevOps culture. My teams, they build it, they run it and they have really great best practices around engineering excellence and operational efficiency. If we have an issue in one of our production environments, my teams are on it, and we have great processes around how we do that. We have a really well established COE.
Anytime there's a customer-impacting issue that happens in one of our production environments, my team's right. COE, which it means correction of errors, we review it as an engineering team every week. I sit down with my teams, we go through operational dashboards, we inspect our metrics. We look at how we're doing across the availability latency scale. We have ongoing scaling targets and scale testing. We're constantly inspecting how are we running the service? Not just how we're building it and how we're building new features, but how we're running it.
We run game days as well, so that we try to break our systems and then see that my team, all my on-calls can recover those systems. One of the things that I really like is we put new people in the team on those game days, because where better for them to learn than we've intentionally broken the system. Get in there and figure out if you can fix it before it's actually fixing something in production. That's really great about Amazon.
Then I would say the other great thing about Amazon and Amazon engineering and the teams that I have, I just love what a high caliber they are and how invested the members of the team are, and how hard they will work to try to make the best service for our customers.
Jeremy: Awesome. Well, listen, I am a huge fan of AWS Lambda and I love what you do. I love what your team is doing. Everything that Amazon is doing for serverless is just amazing. One of the things though that I'd love to talk to you about today, and we could get into the specifics of Lambda itself and how everything works, but you have a couple of great talks. You and Marc Brooker did these talks at re:Invent in 2018 and 2019, getting into the details of Lambda, Lambda Under the Hood, right? Great talks`.
If anybody wants to know exactly how Lambda functions work and how all that stuff works under the hood, definitely go check those out. I will put those in the show notes. What I'd really like to talk to you about today is just this idea of serverless adoption or serverless transformation, because I know AWS talks a lot about how all their internal tools are going serverless, right? Which is pretty cool. Of course there's the external stuff too. There's a lot of adoption from enterprises and small businesses and medium-sized businesses and things like that.
I would love to know the mindset internally. How does AWS take serverless or look at serverless and look at Lambda and use that to build their internal processes? What's the learning on that? How do you keep learning and just keep building with serverless?
Holly: Yeah. This is a really fun topic for me to talk about, and as you might imagine, customers find value in the agility and the operational load or the lighter load on operations that serverless brings. My teams are no different nor our AWS teams or Amazon teams. What we have seen over time is teams across AWS adopt and use serverless. Then my own teams over time have also adopted the serverless architecture and they actually want to use it.
Over time, more and more of the Lambda service, in particular on the control plane, because you don't want circular dependencies in your architecture. So we're really careful about making sure that in early design, when we're saying, "Hey, my team wants to use Lambda, is it okay to use Lambda and serverless?" Because it's building serverless underneath serverless and you have to be careful that you're not doing a bad thing. We're really good about inspecting that in the early design phases.
I've seen more and more of my teams picking up and building control planes on Lambda. In particular, they're using the feature that we launched last year at re:Invent called Provisioned Concurrency. What that does for really high-scale, low-latency services is it gets rid of what people have typically talked about, which is cold starts. Of course, we've done a ton of work over the years to reduce cold starts, but they're still not zero.
We're going to continue to do work on cold starts, but for customers who are super latency-sensitive and need that scale and know that they have low latency all the time, Provisioned Concurrency is a great solution. We have used it within our own services as well.
Jeremy: Right. Is that something now where all AWS teams, when they're thinking about building a new service, that they're going to build that on top of Lambda and do that serverlessly?
Holly: Yeah. One of the ways that people look at it is it's that operational model and where are you sitting in that? Of course Lambda's pretty high. We do a lot more of the shared responsibility on behalf of customers, and so teams like that and they say, "Oh, well, this is going to be easier to operate. We're going to get more agility out of it, so let's go there as a first stop." It's only when they say, "Well, may...
View all episodes
By Jeremy Daly & Rebecca Marshburn
5
2929 ratings
About Holly Mesrobian
Holly Mesrobian is a Board Member at Cascade Public and the Director of Engineering for AWS Lambda. Holly has 25 years of experience in designing, building, and managing globally distributed teams in software development, and more than 15 years as a leader of leaders. She has in-depth experience with building services for builders, and for wireless and broadband carriers; online services for direct to consumer offerings; and commercial shrink-wrapped software. With a double Master’s Degree in Computer and Science and Software Engineering, Holly began her career as a developer before holding leadership positions at companies, like Amazon and RealNetworks, and startup Cozi.
- LinkedIn: https://www.linkedin.com/in/holly-mesrobian-a1b710/
- Under the Hood of AWS Lambda 2019: https://www.youtube.com/watch?v=xmacMfbrG28
- Under the Hood of AWS Lambda 2018: https://www.youtube.com/watch?v=QdzV04T_kec
Watch this episode on YouTube: https://youtu.be/nBYUh7CVUiQ
Transcript
Jeremy: Hi everyone. I'm Jeremy Daly and this is Serverless Chats. Today, I'm chatting with Holly Mesorbian. Hey Holly. Thanks for joining me.
Holly: Hi, thank you for inviting me.
Jeremy: You are the director of engineering for AWS Lambda at Amazon Web Services. Why don't you tell the listeners a bit about your background and what the director of engineering for AWS Lambda does?
Holly: Absolutely. Engineering leaders in Amazon are very technical and I think I fit in that class of leader. I've been in engineering for more than 25 years. The first decade that I was in, I was actually an engineer, and then the last 15 years or so, I've been leading large-scale engineering organizations that are also responsible for 24/7 operations. You think about those, they're called DevOps organizations. That's what I've been doing for quite a while now. The Lambda engineering organization is just like that. In terms of my background and how did I get here?
I have two graduate degrees, computer science and software engineering, and as I referenced lots of time, designing and building systems. One of the things that's really great about AWS and leading the teams here, I reference that DevOps culture. My teams, they build it, they run it and they have really great best practices around engineering excellence and operational efficiency. If we have an issue in one of our production environments, my teams are on it, and we have great processes around how we do that. We have a really well established COE.
Anytime there's a customer-impacting issue that happens in one of our production environments, my team's right. COE, which it means correction of errors, we review it as an engineering team every week. I sit down with my teams, we go through operational dashboards, we inspect our metrics. We look at how we're doing across the availability latency scale. We have ongoing scaling targets and scale testing. We're constantly inspecting how are we running the service? Not just how we're building it and how we're building new features, but how we're running it.
We run game days as well, so that we try to break our systems and then see that my team, all my on-calls can recover those systems. One of the things that I really like is we put new people in the team on those game days, because where better for them to learn than we've intentionally broken the system. Get in there and figure out if you can fix it before it's actually fixing something in production. That's really great about Amazon.
Then I would say the other great thing about Amazon and Amazon engineering and the teams that I have, I just love what a high caliber they are and how invested the members of the team are, and how hard they will work to try to make the best service for our customers.
Jeremy: Awesome. Well, listen, I am a huge fan of AWS Lambda and I love what you do. I love what your team is doing. Everything that Amazon is doing for serverless is just amazing. One of the things though that I'd love to talk to you about today, and we could get into the specifics of Lambda itself and how everything works, but you have a couple of great talks. You and Marc Brooker did these talks at re:Invent in 2018 and 2019, getting into the details of Lambda, Lambda Under the Hood, right? Great talks`.
If anybody wants to know exactly how Lambda functions work and how all that stuff works under the hood, definitely go check those out. I will put those in the show notes. What I'd really like to talk to you about today is just this idea of serverless adoption or serverless transformation, because I know AWS talks a lot about how all their internal tools are going serverless, right? Which is pretty cool. Of course there's the external stuff too. There's a lot of adoption from enterprises and small businesses and medium-sized businesses and things like that.
I would love to know the mindset internally. How does AWS take serverless or look at serverless and look at Lambda and use that to build their internal processes? What's the learning on that? How do you keep learning and just keep building with serverless?
Holly: Yeah. This is a really fun topic for me to talk about, and as you might imagine, customers find value in the agility and the operational load or the lighter load on operations that serverless brings. My teams are no different nor our AWS teams or Amazon teams. What we have seen over time is teams across AWS adopt and use serverless. Then my own teams over time have also adopted the serverless architecture and they actually want to use it.
Over time, more and more of the Lambda service, in particular on the control plane, because you don't want circular dependencies in your architecture. So we're really careful about making sure that in early design, when we're saying, "Hey, my team wants to use Lambda, is it okay to use Lambda and serverless?" Because it's building serverless underneath serverless and you have to be careful that you're not doing a bad thing. We're really good about inspecting that in the early design phases.
I've seen more and more of my teams picking up and building control planes on Lambda. In particular, they're using the feature that we launched last year at re:Invent called Provisioned Concurrency. What that does for really high-scale, low-latency services is it gets rid of what people have typically talked about, which is cold starts. Of course, we've done a ton of work over the years to reduce cold starts, but they're still not zero.
We're going to continue to do work on cold starts, but for customers who are super latency-sensitive and need that scale and know that they have low latency all the time, Provisioned Concurrency is a great solution. We have used it within our own services as well.
Jeremy: Right. Is that something now where all AWS teams, when they're thinking about building a new service, that they're going to build that on top of Lambda and do that serverlessly?
Holly: Yeah. One of the ways that people look at it is it's that operational model and where are you sitting in that? Of course Lambda's pretty high. We do a lot more of the shared responsibility on behalf of customers, and so teams like that and they say, "Oh, well, this is going to be easier to operate. We're going to get more agility out of it, so let's go there as a first stop." It's only when they say, "Well, may...