 Welcome to theCUBE's coverage of KubeCon EU 2024, live from Paris, France. Join hosts Savannah Peterson, Dustin Kirkland and Rob Stratche as they interview some of the brightest minds in cloud-native computing. Coverage of KubeCon cloud-native con is brought to you by Red Hat, CNCF, and its ecosystem partners. The CUBE's coverage of KubeCon EU 2024 begins right now. Good afternoon, Kubernetes fans, and welcome back to Paris. We're here at KubeCon cloud-native con, CNCF's biggest European event. My name's Savannah Peterson, joined by Rob Stratche. Rob, Paris is perfect for us. It's the best in the spring. Yes. Has someone said that before? I don't know. Not yet today. Not yet today. At least not on set. And I'm really excited that we have a veteran guest with us, Daniel from NIO. Daniel, welcome back. Thanks for having me. Yeah. I'm super happy to be back. We love having you. Last time we had you on the show, we were in Chicago. How's Paris treating you? It's great. Just like you were saying, the fact that the cold is away just makes Paris way more enjoyable. And it's overall like the food here is amazing. I'm enjoying every single meal I'm having here. Amen to that. And the calories don't count, right? Yes, they don't count. There's zero calories this week. This is a total free-for-all. There are no rules in Paris, as far as I'm concerned. I'm standing all day here in the booth, so they're just disappearing. Yeah. And also the walk in here, it's like at least 10,000 steps to get here, so. I know, right? But I mean, I think again, you guys are in an interesting spot. Min I.O. is in what's going on, especially with all the talk track, I mean, is around AI. And funny enough, AI runs on data. And you guys are at the heart of data and object storage and really bringing those types of experiences to customers wherever they need their data. And you guys had an announcement. What was the announcement that you guys made? Yeah, big news. The announcement we made last week was about the new enterprise object store that we are announcing and releasing to our enterprise customers. And what this new product is about is pretty much we are making a downstream version of Min I.O. that's optimized for enterprise customers, taking all the learnings that we did from all these customers that actually, because of the rush of the maximum amount of scale of data, now we took all their learnings. For example, some of them were, they were used to running on AWS and AWS was, you know, taking care of all their infrastructure needs. But then they were like, okay, let's take a hundred petabytes out. Let's run out on Min I.O. and let's see how that goes. And then they were like, we like this. So let's go to the exabyte scale. But now of course they have, their setups are thousands and thousands of disk. So finding a failure or being ahead of the failures on such a large amount of drives is like finding a needle in the haystack. So of course, this is where we are building all these new enterprise or big store tools that are optimized for helping you find these problems ahead of time. And not only that, there's also a firewall feature, for example, that where you may also want to do traffic mitigation for your own cluster, you may want to throttle the traffic of your freight limit. There's catalog, if you want to do, we expose a GraphQL API, if you want to do searches across the namespace. And there's also caching and a new brand new KMS, that's a whole new KMS that's also available for the Minio Enterprise or Big Store. It's very exciting. And it's in direct response to customer demand. That is true, yes. And so I know you can't specifically talk about some of the customers we talked about behind the scenes. However, can you tell me, because you're seeing a lot across verticals, can you tell me some trends, are there any spaces where people are adopting AI faster than other spaces? Well, certainly in the cybersecurity and automotive industry, we're seeing that the amounts of data they're dealing with, it's insanely large, right? So, and this is one of the big things that we're seeing is people are completely dropping hard disk drives. You're noticing, you know, we'll go all in BME, right? Even though it's four times price difference, but you make that money back in saving costs of human capital. Because if you are with the cheap hard disk drives, eventually they'll fail. And then you have your whole team on a call, trying to figure out what's wrong, what to replace. That's where all the savings for the hard disk drive goes away. But these massive clusters are coming online that I was telling you about. It's all in BME. And it's pretty much that big, and it's all in BME, first to reduce the maintenance costs and the failure rate, but AI demand that scale and that speed, right? Only in BME can see that. And in BME is already going places that hard disk drives can never go. You will see these 30 terabyte in BME drives now coming available, 60 terabyte just around the corner. It's just size of volume, yes. It's wild, it's wild, yes. Yeah, I mean, and you have to bring the management of all that data into a single plane. I mean, and that's got to be a huge challenge. When your customers come to you, are they, well actually, I mean, your customers all seem to be scaling. Do you feel that the open source community is a big part of driving some of these businesses? Definitely, the open source community, it's the one that's driving a lot of the requirements also for Minayo. Why they go and they test our product right away. They tell us, they say, you know, I'm tasting Minayo on this very obscure type of setup and I found a problem. So we like fixing it for them because maybe we extrapolate and we see that problem may hit some enterprise customers as well. So then we go and optimize it. So open source is crazy important for us. Yeah, and you were a user of Minayo, weren't you? If I recall? Yeah, before, yeah. Before I joined Minayo, I was actually a Minayo user and it was definitely the right tool for the right job. I was training machine learning pipelines AI pipelines pretty much and I needed something where I could actually drop my data and have the engineers pretty much built with the idea that you don't copy the data locally. Back then that's where people are like getting it wrong as well, let me put it this way. This is what people are realizing now. You can no longer build pipelines that copy the data locally and then I train. No, I have this very powerful GPUs coming along. Literally a new one just got announced this week, right? So now, and then I need to be able to stream data as fast as that GPU can crunch it and then discard that and then have the next batch being ready and only object storage can offer that throughput. And I was going to say, even though there and it was talked about earlier today in the keynote about how underutilized a lot of GPUs are and part of that is the storage. And so going to NVMe on-premise or wherever is still cheaper than going to NVMe in a cloud because it's controlling the cost, you can depreciate it. Is that what you're seeing from your customers? Is that they're really leaning into that? That is what we're seeing when the customers are bringing these massive closures, right? And it's like 5,000 drives, 7,000 drives setups. And that's their unique of scale. That means when they're running, when they feel that cluster, the next one coming online is the same size, another 7,000 drives. And it goes and tells you that NVMe is actually, right now the technology is maturing enough that all the enterprise is actually betting harder on it. And that's going to benefit everyone because it's going to drive the price for NVMe down as well. So it's the right time to start getting into the NVMe. And being a former storage geek, I guess I could say, when you talked about some of these enterprise features, I think you were talking about how understanding what's going on and kind of the observability. Can you talk to that? Sure, observability traditionally is this, okay, I have these metrics and it's a weekly line going up and down. And that doesn't really tell you there's something wrong unless you see something like a spiking. But when you have 7,000 drives, where's the problem? I cannot put 7,000 lines, it's useless. I need to tell you, you know, there's a specialized visualization where I can tell you and highlight, you know, these drives is impacting you. These two drives are down. And the overall impact is that this section of your data has a degraded performance. So we want to be able to build these tools so you can practically be like, okay, maybe this week's, next week, we'll send someone down to the data center, replace those NVMe drives, and then keep the operations going. No one will ever find that these drives were actually about to implode. So we want to be on top of that. So that's why we're building these specialized tools. And because you're using erasure and coding underneath the hood, it's still protected. And the data's there, so you can go and do that and move the data automatically, moves around, and you can actually unplug drives and plug new drives in. So I think that helps those organizations be more proactive and keep the speed up. Yeah, we want them to be proactive rather than reactive. Because when they are reactive, by the time they come to us, I have a problem with me now. Something's already happening. Yeah, something's already happening. Something's slow. So we want to be, give them, like, here's a tool. You don't even need to come to me. You can actually watch this yourself and they will tell you there's something going here. Historically, this drive was behaving fine. Now you can see it actually means behaving. Here's the data to prove it. Now it's time to maybe automate this drive. Or maybe it's throwing timeout errors or IO errors. Definitely it's that one drive that you need to RMA. Take it out, RMA it. Mina, you can tolerate working without the drive for an extended period of time, just like you were saying. And then when it gets back, or if you have extra drives, just swap them in, you're done. Back in business. No outage, no sad faces, just being on top of the problem. And that's good for both the businesses and their customers. Yeah, it is true. I mean, a little outage can be a huge problem, especially with some of the customers that you were talking about. What, I mean, this is a huge announcement. I can imagine your customers are really excited. What else is in the pipeline for you looking forward that you can tell us about? So, Mina, the Mina Enterprise Object Store, it's a very exciting announcement because it's bringing all these different facets that other enterprise customers are already enjoying. And now that we are actually consolidating into a product and sharing it with everyone else, now everyone else will see all these features and be like, oh, this can really help me. For example, we were showing the firewall around and they were like, hmm, I wonder if I can, the firewall is like a Lego block that I can actually hook on top of some heuristic system and build a dynamic, on top of some AI, some dynamic scaling of the traffic shaping logic. And they were like, I like this, I can build something on top of this, right? So, when customers see this, they'll get their gears going and they'll try to think of it like, oh, this is a problem I had, I didn't even know I had. And now there's a solution, I'll just click a button, turn it on, and start using it immediately. You just touched on something I want to double down on. In terms of maturity in their journey, because obviously GPU shortages, there's a lot of interesting stuff on the hardware and on the software side. Everyone's trying to have some sort of AI strategy but hasn't necessarily ironed that out. We've had a lot of conversations around that. It sounds like you're getting those wheels turning like you said, how mature would you say most big companies you're engaging with are, how mature is their strategy around AI and how are they going to handle the scale? I mean, that's where I was mentioning that they need to modernize their pipelines and they change their thinking because this expensive hardware, you don't want to sit in idle while things are happening. Oh, I'm unloading the previous algorithm, there's a new algorithm, more data, needs to be loaded, let's wait, no. Right. Crunching, done with the result. Load the next algorithm, start crunching. Right, data is streaming to you. So they need to change that mindset and because now they're seeing I have all this data. Now, and something beautiful about LLMs is that it made it obvious to be different use cases where it's like I have this data, now I can ask questions about this data but I just need to run it through the LLM at very high speed. Whether I'm fine tuning or just doing a rack style approach, I need to be able to pull the data when I need it at very fast speed so that people on the other side of the wire are just sitting there waiting for too long. Do you see a lot, I mean, this being a cloud native, you know, cloud native con, I kind of think cloud native con is subsurping or usurping, not subsurping. Yeah, we're making up new words today, Rob. I'm going new words. I love this for us, let's go, baby. And kind of Kubecon, where Kubernetes, I mean, it's 10 years old, but cloud native apps are still being built and a lot of cloud native apps actually are using, you know, S3 APIs to go. Is, do you see that's where a lot of the organizations that you're working with are building their apps really cloud native? Yes, I'm seeing that. I'm even seeing customers maturing out of bare metal. Right, they're like bare metal, I need, I build another unit of bare metal and now I have way too many servers and I'm only using this for my storage infrastructure so maybe it's time to bring Kubernetes. So they're reaching this conclusion, you need something to automate, manage all of this. Do the rolling updates for me, like simplify operations. So they're naturally being, and they're trying to Kubernetes and seeing as a very mature piece of software now. Everyone's very familiar with it. So then they feel more confidence, it's like, okay, I can deploy this myself, I can manage this myself, it's no longer a pain point. Right, so I see companies maturing that way and it makes sense to them because they already move everything to containers, everything's containerized, everything is gearing towards the object storage APIs because it's like, that's what the clear native way of doing things is. So the industry is catching up now and now the AI is like, okay, you want to do this. AI is really pushing everyone like, now you need to modernize even faster because you might not be able to retrofit an AI paradigm on a legacy style. You won't be able to, it's too much data, the scale is too intense. Is Kubernetes the chosen tool for doing that? Yes. Across the board? I mean, this is naming someone, but I love that blog post by OpenAI, where they say, you know, the limit of Kubernetes cluster was 5,000 nodes, we took it to 10,000. So it's like, they are running and what are they doing with 10,000 nodes? They are most likely training the most popular foundation and model on the planet right now. So it tells you about, you know, to scale operations, you need Kubernetes and OpenAI is the proof. They are proud to blog about it and talk about it, right? It's Kubernetes that's actually empowering that. Yes. Well, so I asked you this question when we were in Chicago and I'm curious to see what your answer is today. Next time we have you on the show, which is obviously now a trend. And you're stuck with us. What do you hope you can say when we're in Salt Lake City that you can't yet say today? That's a really nice trap. I was going to say, tell me, there's your sunglasses right there. I can see it on the tip of your tongue. Whatever you wanted to say right there, tell me. No, it's all right. You can tell me afterwards. You can tell me over a drink tonight. Yeah. I'd most likely tell you, I'll be happy to say that I see more customers going extra scale because it's just in just four months, our customers being multi-petabytes and this is hundreds of petabytes. And first it's like, okay, we see hundreds of petabytes, hundreds of petabytes, but now everyone's going extra by it. And it's what that's in such a short amount of time, everyone migrating to Minio for extra scale is just insane. So I want to be able to tell you, now our number of customers I see is more than less than 10. So now it's like, I want to see tens of hundreds of customers going extra scale. Awesome. So we can have one of those customers on with you next time and tell that story. Daniel, we love you and the Minio team and Jonathan, shout out to Jonathan for always bringing you on. Rob, it is a fantastic pleasure. Always. Sitting in the cockpit with you on this lovely jet plane, wherever we're headed, making up words and doing cool stuff. And thank all of you for tuning in to this fantastic three days of coverage from Paris, France at KubeCon CloudNativeCon. My name's Savannah Peterson. You're watching theCUBE, the leading source for enterprise tech news.