 Welcome back, everyone, to theCUBE's live coverage. I'm John Furrier, Rob Streche here, analyzing, as CUBE hosts, we always do, getting in the stories no matter what it takes from the executives, to the leaders of the foundation, the open source contributors, to engineers. Our next guest is an engineer leading a development team at MINIO, a company we've been covering on theCUBE before. Here, Danielle Valdivia. Thanks for joining us theCUBE. Welcome to theCUBE. Thank you for having me. You're in the Distinguished CUBE alumni list now. Welcome to theCUBE. Thank you. We wanted to get you on to talk about some of the cloud-native open source dynamics that are going on. We saw our container ship moving through the harbor. It was perfect timing, our last interview. You got a lot of different projects here at the open source summit. This is not CUBEcom. This isn't 10,000 people in Europe, or 15,000, 20,000 might show up in Chicago. It's a small crowd here at the open source summit. There's a lot of projects going on, but the right people are here. What are you seeing out here in this world? And I know the CNCF is a big part of this group. How many cloud-native conversations are happening here at this event? So, the main conversation point we're seeing is, okay, we're seeing a big shift in the whole Hadoop ecosystem, moving from legacy HDFS stuff, start implementations, right? That was like the nation open source cradle of data on the open source space. But now that the HDFS ecosystem is sort of collapsing on itself, people are asking, okay, what's next? Where should I be moving my data, right? I want to build an even larger data lake, but turns out I cannot move my HDFS into Kubernetes, just using upstream HDFS. So that's where Minayo comes into place, right? People are asking, okay, I already have all these investment on my Hadoop pipelines, my Spark pipelines, my Presto, and what do I do now? So the answer is, okay, you move into Minayo, right? So Minayo is just building high performance or big storage. You change the scheme on your pipelines, and then you don't have to do anything else, right? Everything just works. And that's where the big conversation and the topics that we're seeing on here. Yeah, one of the top conversations that Rob pointed out yesterday in our summary was, there's a lot of dependencies between projects now, to your point. How has that impacted some of the interop, I won't say interop, good, but the cross-project work? So we've been hard at working convincing people the importance of decoupling compute and storage, right? Coming in from the big data world, people thought, oh, I need to make my compute clusters as big as my storage clusters. So now that we've been spending years convincing them, okay, you don't need to collocate them, right? You can actually decouple them. And now that companies are making that investment, when the time comes to actually replace components, right? So of course, they want to make it the least painful as possible. And that's where the S3A connector on the Fado because it's made it a very clever transition, right? Just change the schema, it just works. Yeah, so help, because I mean, obviously you guys have your open source and where are you really helping in, or playing in the other parts of the ecosystem? Are you contributing back to some of the different projects that are going on here? Or what really is the, where you're leveraging in? From an engineering perspective, obviously. From an in-respective, of course. So we're very interested in making sure that open source, Minayo, works for everyone, not only Minayo, but object storage. So object storage has come out as the standard of consuming storage over the internet. And you see that there's big companies making an effort, for example, when Amazon wrote the S3A adapter, it convinced people to start using HDFS. And for example, we saw that some of the ways, for example, Spark consumes storage is making assumptions that it's still using HDFS. And HDFS has, for example, a rename operation, object storage doesn't have a rename operation. So we wrote, for example, a Spark Committer that actually improves the overall throughput of your Spark pipelines, because not everyone is writing highly optimized Spark pipelines. So we just actually wrote this component, and now we're actually considering collaborating back to upstream. That's just one of the efforts, right? In other projects, like SpyTorch, TensorFlow, Q-Flow, we're actually making sure that object storage is always functional. So we're also contributing back in that area. So you're contributing that back outside of the Minayo pipelines, or product, and back into Spark? Or which group is that going, or project is that going back into the HDFS projects? Spark is managed by Apache. So that's going to that team, right? Every now and then, where we make contributions to other smaller open source projects that we see, because every now and then, someone comes with a brand new database, and they say, this is a brand new database, it works on top of the object storage. And we say, that's great, you just need a couple of tweaks to allow for a custom endpoint, and now you can actually test against Minayo. So we like contributing also to all these other projects. Some of them are small, personal projects, and some are smaller startups coming into the scene. So that's where we like contributing as well. So what you're contributing back, just to break it down for the non-technical people, is really how you are able to kind of simulate or mask out the metadata aspect on top of an open, I guess you can say, an object storage layer? Well, yes. And to some extent, what we're actually contributing is just giving people the ability that you know, yes, maybe Amazon came out with the S3 standard, but you know, there's Minayo, right? So the only thing you need to actually make, your product that already works against AWS S3, to work with Minayo is just to change the endpoint. So some developers forget that, exposing the S3 endpoint, it's important. So we just contribute on that part. It's like, yeah, just do it this way. And now developers can work locally on their laptops, right? Your DevOps engineers can build CI, CD pipelines that test against Minayo, and then production can also run against Minayo, right? So that's where we like contributing as well. Do you see a lot of people that are splitting between Minayo and S3, and having different parts of their data lake in different places? Is that something that's big? That is a big trend that we see our customers doing hybrid deployment, right? And when we say hybrid deployment, we see them across clouds. So we see customers, for example, that like keeping some of their data on premise, right, sensitive data. And they like having the exact same storage API as they have on AWS. But moreover, we start seeing that sometimes they build storage infrastructure across cloud providers because now it's trendy that a cloud provider region goes down for a reason, right? So they like to be protected. So they like deploying across two vendors and having the same storage API makes that so much easier. What's the biggest thing that you're seeing in open source that AI is going to have the most impact for? So in the open source world, I mean, the impact that I'm seeing in the open source world from bringing from the AI, it's all these large language models that speed out code that were trained on open source code, right? So even you look at these models and they say, we didn't train on everything, but we train on the most popular open source products, right? And that taught these models to how to code. So the value of all these people collaborating and making high quality code is actually coming back to being in the models. So when engineers are actually leveraging LLMs to write code, they're inheriting all this knowledge that open source contributors are building. And do you see that right now in the progress? How legit in terms of coding? How much human curation do you see going on? Some people are spreading good code. I mean, in known use cases, you say, build me a website or an e-commerce site, it does that. It's not that. I mean, that's in, you know, but Stack Overflow recently had blocked ChatGPT. I don't know if you saw that, I read that on Hacker News. And very interesting because they had all those questions. They have the prompt engine. Yeah, it's exactly. So the interesting thing about, when you look at LLMs, for example, and they're like generalizing knowledge. So the interesting part is, okay, the prompt is quite important. So you will also see that on GitHub issues, for example, but the logic that lives throughout the code is also important for the model to memorize. So observing how code is properly built to address a certain problem, that's mostly what these LLMs need to look, right? So, yeah. What do you think that KubeCon is going to be like in Chicago? We were just, as you mentioned, we were all together in KubeCon, EU. And then it was packed, 10,000, largest I've ever seen, KubeCon in Europe. That was the biggest we've had. That was pretty bad, 2,000 were on the waiting list. So to me that's a tell sign that the US is going to be a monster show. What do you guys think this shows, what do you think the show is going to be? If 10,000 in Europe, what do you think Chicago will be? 15, 20, what's your guess? Oh, just guess. At least 20. At least 20, yeah. Rob, what do you think? Yeah, I think it'll be, you know, probably closer to 15, but they'll have a wait list after that. I'm sure that'll, but I'm sure it'll be close to 20 wanting to be there. That's for sure. Yeah. All right, so the question we have to ask you as we wrap up this segment, we'll finish this segment out with the platform engineering conversation. What is your view on platform engineering? Because it's become quite the conversation, a movie we've been talking about on theCUBE for a long time and everyone knows that, but it's become more of an IT definition. As DevOps comes more mainstream, platform engineering used to be like Google SREO, yeah, hardcore, spitting glass, eating nails, you know, hardcore developers, right, hyperscalers. But now as platforms go to IT, that you got that platform, you guys do a lot of storage stuff, okay? So, okay, you got platform engineering, you got security, policy data. Now apps are in the cloud or on-premise with cloud operations. They got APIs, they're connected, so they have dependencies, so you got apps have platform-like features too. Yeah. How do you view platform engineering versus just a robust app that, by the way, works together with other apps? Yeah, well, you see, when people are building this transitioning to moving their applications into the cloud and Kubernetes has made a great job in standardizing that, right? Kubernetes has emerged as the operative system of the internet. So, all these applications that people are moving into the cloud, they're writing them under certain assumptions, right? I'm expecting something, some standardized DNS, service discovery, some standardized storage through CSI drivers, and some standardized object storage, right? So, people like their storages like that. So, coming from the legacy world, people used to, okay, I'll buy some appliance, they'll send it to my data center, I'll put it, but if you're moving into the cloud, you don't ship an appliance into AWS data center or Google data center, right? So, you expect also your object storage to be software-defined, and to be able to, that identity department would really like that it's also declarative, right? So, this is where Minayo comes easy, right? So, it's like, okay, I need an object storage for certain business unit, and I need with this capacity, encryption, this certain resiliency capabilities, and that's it, right? It will happen because it's on top of Kubernetes. How important was open source to your company? It was really important, right? Building trust from the community, being like, okay, I understand why this thing is so fast and so reliable, right? Being able to see and then people contributing, right? If we make a mistake, we find out within minutes. So, if we break something in our releases and we make a bad release, right? The community within 30 minutes will be like, this broke, right? So, we'll be like, okay, that's true. We'll patch it and make another release. So, that's the value of open source, right? With open source, when will we know that some obscure pattern header and some API broke? We wouldn't be able, right? So, the community is actually like contributing back to us and helping us hardening our software. Final question, how has cloud native technologies helped you guys be successful? I mean, obviously, you have an alternative to AWS, but cloud native has also been growth with CNCF. How important has that been to your success? So, I think what has made me successful is that all these other cloud native technologies are acknowledging that, okay, I can build now different products that rely on top of object storage. And I need a component that can go along with me, right? Have you seen all of these other projects that say, well, this is how you could do it against object storage? And here's an example with Minio, right? Because they even test their product that way. So, that's how it has been benefiting us, that the open source community also see us as a reliable alternative of, doesn't matter if you're running my product on premise, on this cloud provider, on this other one, as long as you have Kubernetes around, or even if it's very metal, but we have a Minio around, things will just work, right? And this is how we've been seeing it helping us. Daniel, thanks for coming on theCUBE, great to see you. Thank you. Welcome to the first time on theCUBE, Daniel Minio, engineer in the trenches, making code. Are you shifting left, making sure everything's secure? Yes. All right. Security, data, AI, cloud native, open source, all magical pieces of the equation here at theCUBE here in an open source summit. We'll be right back with more after the short break.