 From theCUBE Studios in Palo Alto in Boston, connecting with thought leaders all around the world, this is a CUBE Conversation. Hello everyone, welcome back to our CUBE virtual coverage of able to have summit online, also a virtual event. I'm John Furrier, host of theCUBE. We're here in our studios in Palo Alto with our quarantine crew. The past two and a half months, continuing to keep theCUBE rolling, keeping the lights on, talking to everyone that's out there, also covering the top events. A to the summit, the CUBE can't be there, the event's not happening, we're happening. Virtually, got a great guest here, Keebalon Harain Oberoi, Director of Product Marketing, Database Analytics Blockchain. He understands data, understands infrastructure. Great to see you again. Thanks for coming on, virtual CUBE. Thanks for having us and this is a cool way to do it. Yeah, hey, you know, now there's no excuse when I hit you up on LinkedIn. We're going to do a video, a lot more to do. Anyway, in all seriousness, this is a tough time at scale problems are here. We're seeing more and more things going on. At the summit here, Kendra, general availability, general availability of the ultra warm for Amazon Elasticsearch on a non-augmented AI. A lot of the GAs are coming from re-invent. So a lot of, you know, the cadence of AWS is happening now. You're involved in the Elasticsearch, the ultra warm, this kind of gets to the role of data, warm data, cold data, hot data. This is a big part of the machine learning. Can you give it a set up for why this is getting so popular? What's the big deal here? Yeah, absolutely, yeah. And so I'll start off by just setting some context on Elasticsearch itself and why it's gotten so popular more recently. So, you know, we talk about data and having to grow exponentially over a long period of time. And it's because, you know, so many people are now building apps in the cloud using microservice architectures and the amount of log data that's being generated by these applications is being used to monitor and assess, you know, the operational performance of these systems. And so a lot of customers are moving to Elasticsearch service because it allows customers to collect and analyze and visualize all of this, you know, unstructured and semi-structured log data, this machine generated log data, in order to kind of look at how the applications are doing. And so Elasticsearch service is sort of the fully managed cloud version of the Elasticsearch that allows customers to run Elasticsearch, you know, in a fully managed way, which means I'm not spending time doing configuration and setup or, you know, figuring out scalability for my clusters. I can focus more on actually analyzing the data itself. So that's going to just a little bit of what Elasticsearch is. And so what's been going on is as customers have been using Elasticsearch, the amount of data that they want to be able to analyze is increasing. And so one of the challenges is that Elasticsearch, it's the file format itself. It's really optimized for search. So it makes it really quick and interactive, but it's not optimized for storage. And so it's somewhat inefficient for storage. And so what customers end up doing is if they want to store months of operational data, it's actually hundreds of perabytes. And so what happens is it becomes expensive and customers either start to store that in archives or they don't store it at all. And so if you store it in archives, now you've got DevOps engineers and security experts that have to spend days to restore that data from the archive in order to sort of search and analyze that data. And so what UltraWarm does is it's a new high-performance, low-cost warm storage for Elasticsearch service. And this allows customers to store up to three petabytes of data at about a tenth of the cost of existing options. And so it gives customers the ability to now store months of data for interactive analysis than they could before. That's a great description. Thanks for sharing that. I think one of the things that I've been seeing as a trend, kind of old guard mentality was, hey, here's some storage, you're going to pay for it. Oh, let's do some tiering and pay for that. And then that's cool. But then as you get more and more data, when you said that log files and unstructured data, you need to use that not only to store up but to use it in the applications. The data is actually part of the user experience, right? So I think that's where I see it. Now, what you're saying is that the old model or even the cloud model was getting costly because they were storing the data because they needed low latency. So is warm implying lower latency faster access to data as well? So I get the pricing thing, so it's lower costs, we'll get to that in a second. But is it a speed issue around access to data? Yeah, so ultra warm, so that gives you the best of both. So elastic search, like I said, it's optimized for search so you can get that fast interactive query and visualization of the data, but it's not optimized for storage. So with ultra warm, you now have a warm storage sphere that's sort of optimized for both. You can actually still get that interactive query and visualization capability that you would expect from elastic search, but you can do it at lower costs in a much larger amount of data. What are you talking about in terms of order of magnitude here? Give us a taste for the warm cost structure versus the alternative. Yeah, so it's roughly about 80% lower than warm tier storage from other in a managed elastic search services and you'll get about 50% faster query execution. And so that's enough for customers to be able to get that interactivity they want from that sort of elastic search experience that they're looking for. That's pretty significant numbers there. This has come from the Amazon architecture, Nitro, what's the secret sauce in all of this? Yeah, so ultra warm, it's effectively, it's a distributed cache, and it's a distributed cache for more frequently accessed data. So what it does is it uses these advanced placement techniques to determine specifically which blocks of data are going to be accessed less frequently and it moves those outside of the cache into S3 that's low cost storage and then for the more frequently accessed blocks, you know, it'll keep that in the cache so you can get that interactivity. So it's effectively doing really, really smart caching on really large volumes of data directly inside of elastic search service itself. And the value for me is the customer is what? I've got acts better integration for data intelligence into the app. Is it machine learning? I mean, it's a multitude of problems. You can now do your operational analytics and log analytics on a longer period of time than you would at a much, much lower cost. And so if I'm a firm that's doing analysis of my security logs and I'm only able to do it cost effectively by looking at my security logs for the past week, I can now cost effectively do that same analysis by looking at the security logs for the past month and that might actually give me the ability to identify new trends and new patterns that I wouldn't have seen before. So more usable actionable data for the same price it was before, just on a scale. So more scale for data, making it usable with the application. Yeah, more scale at lower cost. More scale, sounds like the Amazon formula. All right, so what's the most important thing to take away from this cost structure, scale, anything else that we should know about around ultra warm for elastic search? Yeah, I mean, the biggest thing is, you know, again, it's the amount of, it's how your analysis changes. You can now go from storing just kind of a few days, maybe weeks worth of operational data to months of operational data at really low cost. And so with ultra warm now, you can now use elastic search service for a broader set of use cases as well. Talk about the impact in your upon, I want to put you on the spot here for a second around this new reality, right? We're in an at scale crisis. You guys on Amazon are under a lot of pressure to deliver, I talked with the folks from the EC2 group, Matt Garmin came on as well, David Brown, you guys deliver in massive capacity with compute. I got to imagine there's going to be a data opportunity to kind of have more data lakes. I saw the Kendra news in general availability, augmented AI, so data will be killer here, feature for the future. This has to be more ubiquitous in terms of capability. What's your vision on this post pandemic and how do companies reset and reinvent to take advantage of that so that their outcomes are on the up slope post pandemic when it's still going to be a quasi work at home, more teams are going to be distributed. It's a virtualization model of media and life. I mean, we're going to be virtualized. Yeah, I think like everything we do when we think about roadmap, it all starts and stems from working backwards from what our customers are looking for and given the environment now more than ever, moving to the cloud is helping customers lower cost, be more agile, scale up, scale down more effectively. And so it's actually accelerating the need for customers to start to use a lot of data and analytics services in the cloud as well. And as customers continue to look at ways to analyze the applications and how they're running and how to scale the applications, they're going to use a lot of our data and analytics services as well. And so continuing to find ways to give customers better performance, better service and uptime at lower cost will continue to be what we focus on. And we're certainly having those conversations with customers today. And what's your advice to app developers out there and developers who are really going to be on the front lines. The workloads are going to look differently. They're going to have more video, more data. There's going to be more cloud native, more microservices as you pointed out. So how should developers leverage and build great products? What's your best practice? I again, I think for developers just like for us, building great products starts with working backwards from the customer. It's really listening to what are the customer pain points that you're solving, how are you going to solve it in a way that's unique and different, better than how it's been solved today. And then being able to run that in an operationally efficient way that's going to provide a high quality of service in terms of performance, in terms of availability and in terms of cost. All of those things continue to hold true. And our job is to give developers the tools that they need to help them to do that. Well, what else is new with you? How are you doing out there? You had cabinet fever yet? I mean, you got all the tools with Amazon. Everyone's kind of seems like they're in an okay mood. How are you doing? Yeah, no, we're doing good. We're here, I'm here with my family. I have two kids who are doing some version of remote schooling and so juggling time with the kids and balancing that with commitments at work. But here at Amazon, we're kind of very focused on continuing to help customers as they go through this challenging time. And so I think getting the teams aligned on, what can we do to help and getting our teams involved and finding new ways to give customers what they need is the ongoing focus. And we recently released a data lake that's got a lot of information around the whole COVID-19 data sets that are publicly available. And we're starting to see customers use that in particular around the public health space to provide, to do analysis on that data as well. A lot of adivist goodness for you guys doing a lot of tech for good there. Congratulations. Thanks for coming on and sharing the insights. Stay safe. Everyone's got Kevin. Certainly we've got kids, I have four. You know how hard it is. So stay safe and we'll see you soon. And we'll be remote for now. Cube Virtual here with adivist summit 2020 online virtual. I'm John Furrier. Thanks for watching.