Skip to content

Mission Talks: AWS Serverless Session Re:Invent 2020 Insights [Video]

Transcript 

Jonathan LaCour

Welcome back, TIm. Good to see you.


Tim Banks

Thank you, Jonathan. Thank you. Good to be here again.


Jonathan LaCour

Awesome. And again, thanks to everyone who is watching again with us. For the last three weeks we've been putting out lots of videos, having discussions about all the great stuff that's been happening with re:Invent and I just really appreciate everybody. The feedback has been outstanding. We really try to take a more conversational approach to this rather than just throwing recaps at you and I think it's been a lot more fun for all of us. So, perfect.


Tim Banks

Yeah. I'm so happy with this. I feel like, for me, I feel like we saved the best for last on this one, so.


Jonathan LaCour

Oh yeah, for sure. And on that topic - so, reminder: my name is Jonathan LaCour. I'm the CTO here at Mission and I'm joined by Tim Banks who is a consultant with us at Mission. Welcome back, Tim, as I said. And the topic we're going to discuss today is actually the re:Invent serverless leadership session which was super interesting and it was interesting for a number of reasons. We try to do, like, little to no prep for these videos, so I think that that makes it more organic and having good conversation. But, Tim, for me, this is really like Amazon coming out strong and saying, “we're all on the long game here with serverless,”right? 


Tim Banks

Yeah.


Jonathan LaCour

This is an update to our core philosophy about why this is not just the future, but it's actually the present.


Tim Banks

Yeah. They've telegraphed this for a while. If you want to go back to when they first introduced some of the changes, they've had in Lambda functions as far as adding EFS storage, reducing cold start times, increasing the number of languages they have available. The big tell was when they, when they added Lambda functions as part of, like, you know, compute savings plans and things like that. AWS has been moving in this direction for a while. Now, they’ve really just kind of, you know, just pulled the cover off the whole thing to tell you that you know serverless functions and Lambda are going to be the main standard of compute going forward. That's the direction that they've been moving and that's the direction that the industry has been moving, you know, whether silently or loudly, that's the direction we're going. The days of running, you know, a single instance workload are going away.


Jonathan LaCour

Yeah.


Tim Banks

Now, I feel like for the most part we're treating containers like we used to do instances and now serverless functions are the new containers.


Jonathan LaCour

Yep, that's right. Interestingly enough, they're being really smart about how they approach it as well, like pulling down the capability to deploy Lambda functions from container images is just like a, to me, it's like a brilliant move. Now, since the announcement, I've dug in a little bit and there are certainly caveats. But like, there's no question that they're trying to reduce the friction and get everyone there more quickly. I think the biggest tell for me about this was if you looked at the success stories they called upon this year versus the success stories they called upon 2, 3,4 years ago now came to Lambda, this year’s success stories were people like Coca Cola and Capital One, right? Like enterprises, where they're talking about doing things in these very short periods of time with serverless and you would think “oh, that's got to be a startup”, right? Like, no, this is like an enormous enterprise where typically things take forever. And they're embracing it and they're running hard and fast.


Tim Banks

It's pretty cool. Yeah, I thought Coca Cola was so interesting because you talk about how many, how many coke machines are there in the world? You know, and we're talking about changing the way those operate using the agility that you get using Lambda and other serverless functions. I think it was a fantastic story. You know, it tells you that, again, the theme of reinvention, reinventing the way you do business, has been all over this conference. I think that this is one of the one of the best ways to talk about it. Coca Cola, again, a very, very old company. They've been around. One of the first to really be doing the bubbly drink business. It is so ubiquitous that people call you know sodas, you know, Coke, because what you think about. And they're gonna say “hey, it was 2020 things that things fundamentally changed as far as people engaging how they how they get their delicious beverages, so we need to fundamentally change maybe how we do it.” And to get that kind of fundamental change in the short amount of time they did it, leveraging these technologies is so compelling. It is a story that, literally, that they can do it, literally, anybody can do it.


Jonathan LaCour

Yeah. I completely concur. And to me, it's one of those questions where people have often talked about, like “oh, is it ready for prime time? Is this something that can really scale? Is it something we can really use?” If Coca Cola can spend 100 days and transform their entire fleet of, you know, vending machines, to be able to have touchless ordering through an app leveraging Lambda, you betcha it's scalable and it's ready for that time for prime time. Right? Yeah, there's no question.


Tim Banks

AWS has been really good at just taking the complaints, the bugs, the issues that the customer is raising and showing that customer. Just knocking them out of the way, just chipping away at all the roadblocks to adoption that now you have to find. You have to press to find some edge cases where you can't use it as a default compute method.


Jonathan LaCour

I think that'd be valuable to like go through, because one of the things they did is they talked about, they had this slide was like, “What is service about? Well, it's about innovation.” Which I can't stand when people boil things down to just one word like that. It's like, “it's about innovation” - like, what does that even mean? To their credit, they double clicked on it, they went deep and they talked about, like, why, and, you know, if you think about it through the lens of these sort of different stakeholders in a business it really does surface that they're not attacking this just from one angle, they're attacking it from multiple. So for developers, it's about agility and “I just want my code to run, please.” And so you see over the years, they've added things like layers, right, the container images announcement from this year, really ramping up ease of deployment, performance improvements, provision concurrency to reduce cold starts, higher memory allocations - 10 gigabytes was announced this week which is like incredible, and you have to keep in mind things like observability, right, and that starts to get into operations as well.

Tim Banks

Yeah, yeah, yeah. And you know again as an old ops guy, the notion that now we're going to get better insights into how these service functions are running, how they need to be deployed, how they need to be scaled, maybe how they're failing, again it's a lot of concerns about how we get them deployed and how we get them deployed on the fly and very quickly. That speaks to me. Those are pagers that are not going to go off. That is sleep I'm going to get, right, that I wasn't getting before and that's a super compelling story. When I'm looking at trying to adopt some of these things and I'm looking at what the technical debt is going to be to get to that point and realizing that on the other side of that it's going to make our operations run more smoothly, it's going to give us back our lives because they are being changed every time a pager goes off, I mean, that's a compelling story. That's a good way to sell people on the effort it is going to take to adopt this.


Jonathan LaCour

And it's not just about Lambda, right. Like I harp on Lambda because, look, my background is development, not operations. Now, I do a lot of operations and have done it over the years so I'm very excited about that aspect as well but, you know, when you think about above and beyond Lambda, there's a lot of serverless that happens at AWS. Every new managed service and new fully managed service that they're launching is just this incredible burden taken away from you as an operator. Right. You know, I mean, say what you will about Kubernetes, for example. It's a really good example of, like, have you ever operated a Kubernetes cluster before? It sucks. It absolutely sucks. You do not want to do that. That is not a thing you want to do. AWS looks at every single thing, every single open source stack out there, they announced airflow (like a managed airflow) relatively recently, managed streaming Kafka, Aurora Serverless which is their own kind of custom package. The whole thing is, you don't have to deal with anything other than the fact that it's an API, an endpoint that you talk to that does what it says it's going to do. That's it.


Tim Banks

It is funny because we've been hearing that promise that golden egg for so long that “oh, we're gonna make it simpler to talk to this.” AWS, you know, is starting to really deliver on that, where you just need to deploy it and let it run. I remember back in the early days of Elastic Beanstalk and like, “oh, well, you just need to put it in a run” and it is maybe never that easy. It certainly wasn't that easy with Fargate even with containers, but now they are really, they're really doubling down on that and they’re really starting to deliver. I'm excited to see how this pans out, how some of these functions are going to work in production, especially across a broad number of use cases. I think that AWS has shown what they like to do. It's like nothing is ever, nothing is ever finished.


Jonathan LaCour

Yeah.

Tim Banks

You know, they always keep iterating and making things better. The other thing that I really thought was super compelling, alright, for me, was when they got into that - was it like the single millisecond billing?


Johnathon LaCour

Oh yeah.


Tim Banks

Like, you know, as a cost optimization geek too, that's super nice. Hey, you know, because if you can really fine tune your execution, you can save significant amounts of money. Like it rewards you for optimization.


Jonathan LaCour

And I've harped on this one as well before and I think in a prior video, but I'm going to say it again because to me this is one of the most compelling parts about Lambda because it gets deeper than just the benefit to the developer. It's actually creating relationships inside your business that were hard before, right. It's strengthening them. So, for example, I recall many times over my career where I have advocated internally for doing a refactor of something. Having to try to put together an ROI for that has always been one of the most difficult things to do, especially if you're working in, you know, a workload that runs on servers or on prem environments like I’m going to refactor this. Why, to save resources that we've already got sunk cost in? I don't care about that. That's not my issue. Now it's completely going to the other end. You can have developers and engineering teams going to finance and saying “We're going to do a couple sprints optimizing these six functions which are the top six usage functions. If we shave 15% off of their execution time, that translates directly to 15% off of our bill” right?


Tim Banks

Yeah.


Jonathan LaCour

Which is in an incredible partnership that can be created now. There's no - the ROI is clear, right. There's no more argument to have about it.


Tim Banks

And like I said, I like having now more observability into that sack. You can see exactly what you need to do in order to optimize some of these executions, some where some of these roadblocks are where, some of these bottlenecks are, and that insight literally translates into more money to pay more engineers or to invest back into the company, whatever. So being able to have that kind of instrumentation is now, it's not just for operations, not just for reliability, it is for business critical functionality and for cost optimization. The capability which we've never had before.


Jonathan LaCour

Yeah. Completely true. So, interestingly, speaking of money, let's talk a little about the Capital One case study that they brought up.


Tim Banks

Oh, wow. Yeah, I thought it was super cool because they were talking about, “hey, this is how we had been doing it before and now there's a pandemic on and people are hurting and they're suffering they need to contact us and we’ve got four wait times - people can't wait four hours to talk about some of this stuff like it's super important.” So Capital One took that as a call to action and they changed the way they did customer contacts. Again, utilizing service technologies and Lambda and service technologies, including Lambda, to cut these wait times out and change the way they interact with customers to make it much more efficient and to make it much faster to help people who needed help.


Jonathan LaCour

I think they said they got it down from like four hours like 10 minutes, which is just, I mean, that's incredibly cool. That's incredibly cool. You know, you brought up the theme of this year's reinvent has very much been reinvention and that is a great example of like this big megabank right reinventing itself using technology to to drive a better outcome for their customers, which is, it's pretty cool, right? It’s pretty cool.


Tim Banks

Yeah. I really like that. I think one of the things I thought was important about that, and I think we can all see, is because we're talking about something about something as simple as, you know, an IVR or a way that customers call.


Jonathan LaCour

Yeah.


Tim Banks

Which is something a lot of people maybe wouldn't have thought about using any kind of service functions to do, right? You think of very old traditional ways of doing it where you have trunk lines and you have kind of, you know, these big Lucent systems or Cisco Systems to handle phone calls. But now there are so many more options and especially if you're a larger enterprise that maybe has one of the older legacy systems. it gives you a good idea of maybe how to rethink how you do that because now you have the capability of doing it much more agility, much more securely, and certainly much more cost efficiently than ever before. At a time, especially in times like now, that translates into big changes - big changes in a business and a big difference for people. Like you said before, a 15% cost optimization and your bill that's, you know, several engineers that get to keep their jobs.


Johnathon LaCour

Yep. Exactly.


Tim Banks

Or that's a bonus, you can pay out. That's a real thing that's important to people. We're not just talking about nickels and dimes and theoretical things that make a business, we're talking about things that can make a difference in people's lives. It may seem like, “oh, well, you know, Tim, why would you ever say that, we're talking about tech” but it's what it is. You're utilizing tech to make a difference in people's lives. It's really cool to see how the innovation that these engineers and the folks at AWS and these product teams are doing can make that difference because of the things that they created, because of the ways that these companies are saying, “hey, well we can utilize this.” You're talking about people that have time back in their lives because they don’t have to spend four hours on hold to Capital One. They spend 10 minutes, they get their business done and now they can go about the rest of their day.


Jonathan LaCour

Yeah. And if I had to summarize this entire leadership session in one phrase it would be kind of like, “what are you waiting for?” There have been plenty of reasons to wait on looking into serverless and seeing if it's ready but, you know, between Coca Cola and Capital One, there are customers that are very likely much bigger than you and much more complex than you, that are already leveraging serverless to great success. So what are you waiting for? Are you waiting on more memory allocation? We've got that. Are you waiting for support for building it with common tooling? Well, hey, container images can now be deployed as Lambdas. Are you looking for the right programming language to be supported? Guess what - you could build your own custom runtimes and, by the way, our library of run times is huge now. And it's just like more and more and more.


Tim Banks

“We run state for workloads.” Well, we can handle that too. They've been just chipping away at the roadblocks to adoption. I think what's important, one thing I want to call to mind - we here at Mission, we help people get to that point where they want to be. Whether it's through our insight, whether it's our experience, whether you need us to lay hands on keyboards, whether you need us to be there when you whiteboard, whether you need us to help you do cost optimization. That kind of digital transformation is something that we do over and over and over again. We're happy to help folks get to that point where they can make that kind of difference.


Jonathan LaCour

There you go. And I love it. I think that's a good place for us to close. What a great session this was. Inspiring to think about what can come next for all of us and our customers. I think a couple of quick notes: if you enjoyed this video, you should check out our blog at www.missioncloud.com and you can see all the other videos we've done, including several that Tim and I have done together, a couple with some other groups. We've done them on containers and security, all sorts of good stuff. So check that out for sure. And thank you so much for watching. Look forward to hearing from you and look forward to doing this again next year.


Tim Banks

Thanks, everybody. Appreciate it. 


Jonathan LaCour

Thanks, Tim.


Author Spotlight:

Mission

Keep Up To Date With AWS News

Stay up to date with the latest AWS services, latest architecture, cloud-native solutions and more.