Exploring MinIO’s Innovative Approach to Data Management | KubeCon SLC 2024

This is Textron tv. Hey, everyone. We're back here live at CubeCon, and I'm really happy to interview, uh, to interview and, but to introduce my next guest.

His name is AB Perry. Swami AB was on Textron tv. Oh, it's gotta be three, four, maybe more months ago, right?

Yeah. Um, he's the co CEO of a company called Min io. We're going to hear more about that.

Ab welcome. Thank you. Glad to have you here in person.

Live at CubeCon in the middle of this whole chaos. It's wonderful To Be here. Yes.

Mm-Hmm. So let's talk a little bit about you. I mentioned your CO CEO.

Yeah. That's not a usual kind of Yeah. Situation.

Mm-Hmm. How did, how did it come about as co-CEOs at MIN io? So, Uh, I'm also a co-founder, and when you're a co-founder, titles don't matter.

Right. Especially early on. Yeah.

Right. Like a, I think, uh, I think it's not just true early on. Look at the best performing companies around.

They're still run founder run. Yes. Right.

I think when the founder leaves, That's a very tricky Yeah. It's like making an omelet. You gotta break some eggs.

Exactly. And then some eggs get growth in that. Yeah, Absolutely.

Right. You, you have to be entrepreneurial. The market is so, so dynamic and fast.

You need to be entrepreneurial to actually go grow the business. I'm going to be here as long as it takes, and titles don't really matter. I will, I I have to do whatever it takes.

I will even pick up garbage so my team can do the work. As any founder who does this is worth their salt nose. Right.

Exactly. Very cool. Um, min io Mm-Hmm.

I'm sure there's a lot of people out here. They don't, they're not familiar. Explain to them the MIN io business.

So, MIN IO is an object store. If you look into the public cloud, like you have cloud storage. When your data grows and when your data is sensitive at scale, particularly the AI data, you are going to go to private cloud.

When you go to private cloud, you are going, you're not going back to the traditional San NA and virtual machines. You want your private cloud to look exactly like public cloud. You're in the public cloud.

Sure. VO runs there. You can use AWS S3 or Google Cloud storage when you go to private cloud.

We are pretty, we are the only one who can scale at that level. And, uh, nowadays we are seeing, uh, all the AI deployments have reached multiple access scale. It's just so crazy.

Right. And mean is basically a storage system for storing your AI data. Got it.

Now sounds easy. Mm-Hmm. Of course.

The world between AI and video and just everything that's going on in the world of technology. Yeah. Today, the, the scalability of the solutions Mm-Hmm.

That are, is required for these Mm-Hmm. For today's use cases. It's crazy.

It is. It's, I mean it's, it's, it's almost, you know, like, it's like when, you know how we talk about planets and suns Yeah. And you really can't get your head around 500 light years.

Yeah. I'm not saying we're at that, but we're getting near those kinds Yeah. Of numbers in, in storage.

Yeah. Access. Give us an idea.

I mean, you are rather humble and modest. Mm-Hmm. But talk to us, you know, what are the use cases where, yeah.

MIN io is really kind of shining these days. So the, the open source side, we are everywhere. Use cases, I cannot imagine.

Right. Like the product was never designed for those cases because the product is so versatile, so simple, so like, so efficient, it just runs everywhere. But all that success led us to Minio becoming an industry standard.

We are the single largest player in the object store space by adoption. But as a business, we want to be picky about the use case that we want to go after. Increasingly, it's now all our com com commercial successes coming from the AI data space.

And the reason being that the size is so large, this is a problem that we are uniquely positioned to solve. So the commercial use case, I don't want it to be many use case when it comes to open source, we want to be everywhere because you want to create an industry standard when it comes to commercial journey. We are very picky.

And the, the use case that we are going after is AI and data database, data processing. If you put data at the heart of your business, we want to be the system where all your data is. So you can do large scale AI and data workloads on us.

So the com the use case that we are really focused is mini, is AI use case. The use case where Miao shines, it's all over the map. We kind of boiled the ocean.

We needed to do that to make object store and industry standard. It's a new technology that born in the cloud and now it has to reach everybody. So the use case is all over.

But let's, you know, one, I, I didn't do this always right. I had started a bunch of Mm-Hmm. Product companies co-founded.

And you know, one of the things I learned in running product was you can't, you gotta remember what your corner cases are Yes. Versus your main cases. Yeah.

And if you are just developing a product strictly for those corner cases Mm-Hmm mm-Hmm. You're a loser. Right.

Absolute. You've gotta go after the mainstream. Yeah.

Sometimes you just gotta walk away and say no to the corner case. 'cause it's not what you're designed for. Exactly.

But in today's world, what we used to maybe think of as a corner case is now, especially when it comes to usage. Mm-Hmm. And scalability.

Mm-Hmm. It's now the every day. Yeah.

If you wouldn't mind share with our audience kind of your Mm-Hmm. The, let's call it the highly, highly scalable mm-Hmm. Kind of use cases that you are encountering.

Yeah. They kind of fall under the hyperscalers. Maybe they're level two, level three hyperscalers.

Probably they don't like to admit it that way. End of the day, that once you cross a hundred petabytes, a hundred petabytes, nowadays they actually call it a scalable unit. But these large scale customer of ours, a hundred petabytes used to be gigantic.

Just few Years. It still is gigantic. Yeah.

But go ahead. In the gene I world is completely different. Right.

A hundred petabyte is, is baby cluster. Jonathan would like to joke it as baby pod Uhhuh. And the, and the scale now is multiple exabytes and multiple exabytes.

When you are sitting in the public cloud, they absolutely allow the public cloud infrastructure. They allow the convenience. You have unlimited budget, I would tell you just stay in the public cloud.

But the nowadays modern businesses, the unit economics drives everything. And also the scale of AI starting point only is so large. They know how fast it is going to grow.

They are now capturing more data than ever. And the data is unstructured data, audio video is multiple times larger than at a row or a table in a database. When you, when this data grows, they're anticipating a tsunami of data that's going to drown the organization at this point.

That's where customers bring in M io and they, uh, the, the scale, uh, the, the, they kind of, they put them under hyperscaler and the customer behavior has changed as well. When you are small scale, you traditionally bought an appliance completely integrated SAN or NASA and VMs. You can't do that.

In the AI world, the AI world, the entire infrastructure has been ripped apart and put, put together completely differently, is no longer CPUs. The heart of your compute is GPU, the network a hundred gigabit is too slow. We are talking about a hundred gigabit per node.

Each node having dual a hundred gigabit is too slow. 400 gigabit is the new starting point. Now, even that is a bottleneck.

The thing is that why do I need this? Why do I care? If you don't put AI at the heart of your business, you're going to miss out very quickly.

The industry is transforming so fast. The companies that adapt, adapt to this new environment is going to succeed. And AI has already shown the promise it can do to your business, particularly the enterprise business.

That data is their core asset. AI helps them unlock value. So this is not a choice anymore.

Right? Absolutely. So sometimes numbers matter though.

Share some numbers of some of the bigger, uh, installs and customer, you don't have to name names, but Yeah. Give us an idea what we're talking about. Yeah.

Uh, these are basically a like, like in exabyte and multiple exabytes. And, uh, they are now, some of them, uh, actually are now starting at 10, are now starting to look at, I need to plan for 10 exabytes. The enterprise market.

Uh, even the, even the commercial banks, they were at a hundred petabytes and then that grew to 500, 600 petabytes even during the Hadoop and analytics era. They, they, they're pretty much, if you talk around like commercial, like all the major banks and FSI segment, they would, they would tell you that they already have 500 plus petabytes. The problem they have now is they need to plan multiple exabytes.

Now, if they're, if they're going to modernize it, they are modernizing it for ai because they now you can see, uh, for all the large enterprises, exabyte is kind of the new starting point. And if you are, uh, if, if, if you only have 10, 20 petabytes of data for you to play a big role in the AI market, look at, look at who's going to win in this market. Not the ones who, or who has the largest GPU infrastructure.

The ones who has the most data, the companies who has most data, most are figured out how to unlock value of the data, are the one going to win. Tech is easy to acquire. Data is not.

Data is your property. Yeah. Everyone can acquire tech, but not everyone can acquire data.

Well, So more data, more, more money. And that's the scale. Unless you're one of these people who just took like a whole copy of the internet Yeah.

They that your data. But that's another story. Yep.

You know, it's interesting you said that though, because I actually interviewed someone earlier. So they, they're offering, let's call it GPU as a service. Mm-Hmm.

mm-Hmm. So it really is just hardware is service. Look and the time, time where you can't get a lot of GPUs.

Yep. It's a commodity, but it, it very quickly becomes a commodity. It is.

It is. And, and so that's not a business one wants to get into. Um, on the other hand though, is there a limit?

Like how much data is enough already? I I think if you ask the customer, they'll tell you it's never enough. Because data is like, the more information you can extract, it's not like it, it depletes compute.

Once you compute Right. Like a, a, that computation is done. The problem, the advantage of the data, the more you accumulate it, the, the valuable information you can pull out of it, it compounds very quickly.

The bigger data has whole new, like, emerging behavior. It has whole new insights. So the more data means it is better.

And also when you take information out, you don't deplete it. The more you make, it's just like making software copies. The data keeps on paying you.

The, a simple way to understand this, Instagram, WhatsApp, these kind of tools, right? TikTok, the amount of money they're spending on their infrastructure is insane. But how can it be possibly free?

You don't even pay a dollar a month, but not a dollar a year. Someone's gotta pay for this stuff data. Right.

And that is the behavior now is, uh, percolating down into the enterprise. Data will not only pay for all of the infrastructure multiple times put together, is going to power your business. If that was not, if look at the consumer world, it's already proven.

And that's now coming into the enterprise and AI has just shown them the way. So does there, does there come a point of pushback though? Mm-Hmm.

Like, we're seeing this, we mentioned before, you know, some of the AI models, let's say borrowed liberally Mm-Hmm. From other people's data. Mm-Hmm.

Because it was available on the internet. Mm-Hmm. But you still own that data.

Mm-Hmm. So has there come a point where it's like diminishing returns? I I think it is more, it's not from the data point of view data.

There is always more value. And, uh, there is like, there are different kinds of information you can pull out of the data. The models are on the other hand that they, they, they're evolving.

They are still in our in ancy. Even what is out there, just if you took the open source models alone, you can unlock incredible value out of your data. So, uh, so should I wait for larger and larger models to come the, we are already at a point where we can start showing Im impact on the business and the next phase, every six months is going to just move so fast.

I I, I, I actually don't see the problem of that. There is a diminishing return. If that's the case, you would've seen this with all the meta family, all the consumer TikTok, they would stop taking data, right?

They, they're not. Instead more data is only getting better and better. Why are those companies so powerful, so well positioned to become the next generation of mega super mega brands?

What differentiates them? What is their mode? That data is unique, property of theirs.

What enterprises are now starting to realize the questions you can ask on a public LLMA, any GPT, anybody can ask. It's trained on a generic world knowledge. There is no point in you trying to capture, download the internet.

But what you have for your business, GPT cannot answer is unique property of yours. More data that is unique to your business that you can unlock. There is no end to it.

It, it, the more and more data is going to get better and better. I love it. Hey, beware out time.

They're telling us we had more to go over. But most importantly, for people who wanna get more information about the MIN io object store website. Yes.

io. Let me ask you a question about that. Mm-Hmm.

Word is that they, they're doing away with the IO top level domain. The company, the country that was the io that island in the Indian Ocean, uh, uh, became part of March or something. Yeah.

Yeah. What do you, what do you hear? Are you worried?

What do you think? I think it's really hard for them to let go. Uh, it's a a but it doesn't matter to us.

We have main AI as well. Right. It's a end of the day.

Even if you see how people discover, uh, the websites, they just go to search engine and then the browser takes you. Right. Even if they, if they drop the top level domain, do io, there is going to be a transition period where it will redirect.

And I like if the world changes, it's not just alone. We are alone. Suffering IO was the, was the TLD of choice for majority of the companies.

Right. com. Right.

And it is not us If the rest of them are going through the same journey, I'm okay with, I, I just look, I think it's a huge money maker. It is. Why would you walk away from that revenue, but Exactly.

Whatever. Anyway, it, it's a pleasure to have you here in person. Yes.

io here at, uh, cube Conn. We're gonna be back in just a moment. Stay tuned.