 from the Regency Center in San Francisco, it's theCUBE. Covering Serverless Conf San Francisco 2018, brought to you by SiliconANGLE Media. Hi, I'm Stu Miniman and you're watching theCUBE here at Serverless Conf 2018 in San Francisco. Happy to welcome to the program Michael Garski, who's the director of platform engineering at Fender. Thanks so much for joining me. Thanks, Stu, thanks for having me on. All right, so, luckily I don't need to introduce Fender because I think most of our audience would be familiar with Fender, guitars, music, all that stuff, but we're going to talk a little bit about the tech side. Okay. Even though, let me ask, there's a question I usually ask. Most companies, going through the digital transformation, lots of changes there, how does digital impact Fender? Fender started a digital division in late 2015 and it was a focus on all new digital products to complement our well-known physical products. And since then, we've launched Fender Mod Shop where a user can order a customized guitar online, have it delivered in a month. We have a single sign-on solution across all of our web and mobile applications, a guitar tuner, we have connected amplifiers with companion apps to control amplifiers remotely, and our flagship product is Fender Play, which is an instructional app which allows a user to learn to play guitar, ukulele, and coming soon, bass. Love it. I'm guessing that has something to do with what you're involved with on the cloud and serverless side to enable those technologies on the mobile side. Exactly, we've fully embraced AWS Lambda to support all of the services for the web and mobile applications. Okay, so A Cloud Guru is a training company and we've talked to them extensively as to why serverless was a good fit for them and enable them to do it, but bring us through what led to your adoption of AWS Lambda. Give us a little bit about what kind of technologies you were using before and how you got to this solution. Well, we started out building services and go to standard EC2-based microservices, and then we started tinkering a bit with Lambda. We had to ingest a product catalog from SAP. They could deliver a file, drop it into an S3 bucket, so it was simple enough to create a function that can process that file and ingest it into Elasticsearch. From there, we used custom authorizers with API gateway mapping templates to save custom tunings for users. And then as we started building for Tone and Play, Tone especially is a very limited audience. It's whoever buys the amplifier. So we're not talking millions of people, it's gonna be hundreds of thousands. So it was a very good use case to go ahead and do that. And same thing with Play, when you were starting a new business, that's a great model for us that we can just pay for usage. All right, so yeah, it sounds like you were using cloud and they've used its model fit for what Serverless was built for, correct? Exactly, yes. All right, how much is management aware of kind of the underlying technology? Is your group kind of just allowed for free reign to kind of deal with this? Or are there anything you need to go to the CFO and be like, well, are billings going to change a little bit compared to what you might have known in the past? It's, we pretty much have free reign. And our biggest AWS expense is actually what we pay for an AWS Glacier for storage for the raw footage of all the 4K footage from the instructional video shoots. And Lambda on top of that is simply a rounding error. Yeah, excellent. And the mobile apps that you've built, is that a, are there trials on there? Is everybody up to sign up to be able to use it? Is it a freemium model or is it a paid model? The tuner is a completely free application. There is an in-app purchase for like board and scale libraries and some pro features of the tuner. Custom tunings are free. The play application comes with a 30 day free trial. So a user can sign up either on the web or via the Google or Apple app stores on their mobile device. Okay, so with that kind of model, I would think that Lambda would be nice. There's, you know, you said your expenses aren't that high using this kind of service. No, not at all. Like in the month of June, we spent, I think it was like $132 for 68 million Lambda invocations. And to kind of put that in perspective of it, it's what we pay for some EC2 services, EC2 instances that support our legacy authentication service. But we're also moving that over to Cognito now so we can get rid of all the EC2 instances. Okay, when you started using this technology, how'd you first learn about it? How'd you get up to speed on it? Tell us a little bit about kind of training adoption. There's a lot of experimentation. So we have it set up where we use one account for our QA and production environment and another account for our development environment. All the engineers on the team have free reign to do whatever they want to in the development environment and spin up whatever they need to. So we just started playing around with things and experimenting like let's hook up Lambda functions, API gateway, oh, this is going to work really well and just kind of proceeded down that path. All right, great. Any learnings, anything that you've tried playing with and said like, wow, this just isn't going to be a fit for what I need? Tell us what worked, what didn't. I would say about the only thing we found that really doesn't fit within Lambda and serverless would be really very low latency applications. Like you're doing an autocomplete for a search system. You want that snappy. It's, humans observe, I think it's about a hundred milliseconds, things seem instantaneous. And that's going to be very challenging to get from API gateway Lambda to get that consistently. Great, and you're speaking here at the conference. How that ended up happening? What are you looking to share with your peers? I happened, I was, I submitted a talk for a conference and then Drew from A Cloud Guru approached me and asked me to submit and to tell him I already did. So they went ahead and approved it. And I'm sharing what we've done and built at Fender Digital and sort of what we found as far as tools for monitoring performance optimization as well as some things to really be cautious of when you're dealing with Lambda, especially with regards to concurrency controls. Great, and how have you found the show so far? You were at the keynote, got about 500 people here. Yeah, it's really interesting. I'd really like to focus all on serverless. You see, go to a lot of other conferences there might be one or two talks that kind of focus on that. It's nice to have something completely focused in that space. All right, and from a maturity standpoint, are there things that you're looking for in the roadmap from Amazon? They've been baking serverless kind of into all of their services. So do you expect to stay on Lambda or are there other services that kind of fazz or serverless built into it that you might be using? We expect to stay on Lambda for the near term. I don't, we don't have any plans or looking at anything else like the Azure or Google Cloud functions. Our intention is to stay with AWS. They have a lot of other services. They're new machine learning services. We use DynamoDB quite extensively and so we're probably going to stick with them. Yeah, but it's inside Amazon. They've been expanding their serverless portfolio as it was. Oh, yeah. And I remember when I was at the show when Lambda was announced and then it's Aurora with serverless underneath all those. So do you expect to adopt some of those other services that have AWS serverless kind of baked into it as opposed to just using kind of the Lambda tool? Absolutely. Especially, we just mentioned the Aurora serverless model. That's one that we're taking a look at and evaluating is we've got some data in DynamoDB but as requirements have shifted in the business over time it's really become very difficult to model in DynamoDB. So we're going to kind of take a look at that and possibly move to Aurora serverless. Yeah, I'm curious, you know, how does Fender think of the data involved? Is that something that you mentioned AI? Some of these, is that something you'll be able to take the data and leverage that potentially even make new business revenue streams out of that in the future? We're doing some of that already by just watching user analyzing user behavior so we can improve our products internally. And we're looking at adding more features to where we can really understand what people are doing and then make our products better. All right. Michael, want to give you the final word for your peers out there that might be saying, hey, I've heard of serverless. I'm kind of thinking at it. What advice would you give them? Just dive in, get started. Don't hesitate. It doesn't cost you anything really to experiment with it. That model works very, very nice. Yeah. And it's one of the things that's great. It used to be you would take a lot of peer-to-time and some big investment to be able to try a technology out or maybe you'd get some demo but serverless is pretty easy to get started. Exactly. Especially if you're using a framework like say serverless framework or maybe using AWS, excuse me, AWS's serverless application model that really helps as far as setting up all the resources that your function needs as well. Well, Michael, really appreciate you ripping with us on your deployments with serverless and I hope your peers will definitely check it out. All right. Lots more coverage here from the serverless conference here in San Francisco. I'm Stu Miniman and thanks for watching theCUBE.