 Welcome back to theCUBE's continuous coverage of AWS Storage Day. We're in a beautiful downtown Seattle in the great Northwest. My name is Dave Vellante and we're going to talk about file systems. File systems are really tricky and making those file systems elastic is even harder. They've got a long history of serving a variety. Use cases with me is Duncan Lennox, who's the general manager of Amazon Elastic File System. Duncan, good to see you again. Dave, good to see you. So, tell me more around, specifically the Amazon's Elastic File System, EFS. You got a broad file portfolio, but let's narrow in on that. What do we need to know? Sure. Yeah, well, Amazon Elastic File System, or EFS as we call it, is our simple serverless set and forget elastic file system service. So, what we mean by that is we deliver something that's extremely simple for customers to use. There's not a lot of knobs and levers they need to turn or pull to make it work or manage it on an ongoing basis. The serverless part of it is there's absolutely no infrastructure for customers to manage. We handle that entirely for them. The elastic part then is the file system automatically grows and shrinks as they add and delete data. So, they never have to provision storage or risk running out of storage and they pay only for the storage they're actually using. What are the sort of use cases and workloads that you see EFS supporting? Yeah, we built EFS to support a broad set of customer workloads. So, it's everything from, you know, serial, highly latency sensitive applications that customers might be running on-prem today and want to move to the AWS cloud up to massively parallel scale out workloads that they have as well. So, okay, are there any industry patterns that you see around that? Are there other industries that sort of lean in more or is it more across the board? We definitely see it across the board, although I'd have to say that we see a lot of adoption within compliance and regulated industries and a lot of that is because of not only our simplicity but the high levels of availability and durability that we bring to the file system as well. The data is designed for 11 nines of durability. So, essentially you don't need to be worrying about anything happening to your data and it's a regional service, meaning that your file system is available from all availability zones in a particular region for high availability. So, as part of Storage Day, I think we saw some new tiering announcements. What can you tell us about those? Yeah, we're super excited to be announcing EFS intelligent tiering and this is a capability that we're bringing to EFS that allows customers to automatically get the best of both worlds and get cost optimization for their workloads and how it works is the customer can select using our life cycle management capability, a policy for how long they want their data to remain active in one of our active storage classes, seven days for example or 30 days. And what we do is we automatically monitor every access to every file they have and if we see no access to a file for their policy period like seven days or 30 days, we automatically and transparently move that file to one of our cost optimized storage classes. So, they can save up to 92% on their storage costs. One of the really cool things about intelligent tiering then is if that data ever becomes active again and their workload or their application or their users need to access it, it's automatically moved back to a performance optimized storage class and this is all completely transparent to their applications and users. So, how does that work? Are you using some kind of machine intelligence to sort of monitor things and just learn over time? And what if my policy, what if I don't get it quite right or maybe I have some quarter end or maybe twice a year, I need access to that. Can the system help me figure that out? Yeah, part of the beauty of it is you don't need to know how your application or workload is accessing the file system or worry about those access patterns changing. So, we'll take care of monitoring every access to every file and move the file either to the cost optimized storage class or back to the performance optimized class as needed by your application. And that optimized storage class is again selected by the system? I don't have to really do anything. That's right, it's completely transparent. So, we'll take care of that for you. So, you'll set the policy by which you want active data to be moved to the infrequent access cost optimized storage class like 30 or seven days and then you can set a policy that says if that data is ever touched again to move it back to the performance optimized storage class. So, that's then all happened automatically by the service on our side. You don't need to do anything. So, you say it's serverless, which means what? I don't have to provision any compute infrastructure? That's right. What you get is an endpoint, the ability to mount your file system using NFS or you can also mount your file system from any of our compute services in AWS. So, not only directly on an EC2 instance but also from our serverless compute models like AWS Lambda and Fargate and from our container services like ECS and EKS and all of the infrastructure is completely managed by us. You don't see it. You don't need to worry about it. We scale it automatically for you. What was the catalyst for all this? I mean, you know, you're going to tell me it's customers but maybe you could give me some insight and add some color or like what, if when you decoded sort of what the customers were saying is you get inputs from a lot of different places. You know, when you had to put that together and shape it, tell us, take us inside that sort of how you came to where you are today. Well, you know, I guess at the end of the day when you think about storage and particularly file system storage, customers always want more performance and they want lower cost. So we're constantly optimizing on both of those dimensions. How can we find a way to deliver more value and lower cost to customers but also meet the performance needs that their workloads have? And what we found in talking to customers, particularly the customers that EFS targets, they are application administrators. They're DevOps practitioners. They're data scientists. They have a job they want to do. They're not typically storage specialists. They don't want to have to know or learn a lot about the bowels of storage architecture and how to optimize for what their applications need. They want to focus on solving the business problems. They're focused on whatever those are. So you're meaning, for instance, so you talk, tiering is obvious. You're tiering to lower cost storage serverless. I'm not provisioning, you know, servers myself, the system I'm just paying for what I use. The elasticity's a factor. So I'm not having to over provision. And I think I'm hearing, I don't have to spend my time turning knobs. You've talked about that before because I don't know how much time has spent, you know, tuning systems, but it's got to be at least 15 to 20% of a storage admin's time. You're eliminating that as well. Is that what you mean by sort of cost optimization? Absolutely. So we're providing the scale of capacity, of performance that customer applications need as they needed without the customer needing to know exactly how to configure the service to get what they need. We're dealing with changing workloads and changing access patterns and we're optimizing their storage costs as at the same time. When you guys step back, you get to the whiteboard out and say, okay, what's the North Star that you're working? Because you know, you set the North Star, you don't want to keep revisiting that, right? This is, we're moving in this direction. How we get there might change, but what's your North Star? Where do you see the future of this business? Yeah, it's really all about delivering simple file system storage that just works. And that sounds really easy, but there's a lot of nuance and complexity behind it. But customers don't want to have to worry about how it works. They just need it to work. And our goal is to deliver that for a super broad cross-section of application so that customers don't need to worry about how they performance tune or how they cost optimize. We deliver that value for them. Yeah, so I'm going to actually follow up on that because I feel like, you know, when you listen to Werner Vogel's talk, he takes you inside, it's a plumbing sometimes. So what is the, what is that? I mean, you're right. It sounds simple, but it's not. And as I said up front, file systems getting that right is really, really challenging. So technically what's the challenge? Is it doing this at scale and having some consistent experience for customers? There's always a challenge to doing what we do at scale. I mean, the elasticity is something that we provide to our customers, but ultimately we have to take their data as bits and put them into atoms at some point. So we're managing infrastructure on the back end to support that. And we also have to do that in a way that delivers something that's cost effective for customers. So there's a balance and a natural tension there between things like elasticity and simplicity, performance cost, availability and durability and getting that balance right and being able to cover the maximum cross-section of all those things. So for the widest set of workloads, we see that as our job and we're delivering value when we're doing that for our customers. In automation of course is a big part of that. And of course when we talk about, you know, the taking away the need for tuning, but you got to get it right. I mean, you can't optimize for every single use case, right? But you can give great granularity to allow those use cases to be supported. And that seems to be sort of the balancing act that you guys play so well. Absolutely. It's focused on being a general purpose file system that's going to work for a broad cross-section of applications and workloads. Right, right. And that's what customers want. You know, generally speaking, you go after that metal. Duncan, I'll give you the last word. I just encourage people to come and try out EFS. It's as simple as a single click in our console to create a file system and get started. So come give it a try. Easy button. Duncan, thanks so much for coming back to theCUBE. It's great to see you again. Thanks, Dave. All right, and keep it right there for more great content from AWS Storage Day from Seattle.