Data Centers Everywhere – Matthew Trifiro, Vapor IO

This is texturing TV. Hey guys. Thanks for the throw.

We're here with Matt chaffero. Who is CMO for vaporio? And we're talking about kinetic grids and you may well wonder what that's all about.

So, I'm gonna let Matt explain that but there's at least 32 of them now and climbing right man. Yeah, that's right. That's right.

So the kinetic writers name for our product It's a combination of networking and data centers that are built into metropolitan areas. You think about like a city the size of Los Angeles, we build data centers that are spread across the metropolitan area. We connect them with high-speed fiber.

We then allow the full configuration of that for you know, stop losing our software defined networking and our interconnection Fabric and the whole purpose of this is to deliver new classes of workloads to the Enterprise consumers municipalities the way to think about it is any workload that today is locked on premises can now be moved off of premises and delivered from the infrastructure with the same privacy security and performance. As on-premises but with now the cloud economics the consumption-based pricing that sort of stuff we work with all the cloud providers. We work with the we're you know Cloud neutral.

We're carrier neutral. So we're one of the few places at the edge where those two entities can meet. And as you mentioned we're we're customer ready in 32 cities and what that means is we're live in six.

So the six cities were live in our Atlanta Pittsburgh Chicago Dallas Phoenix and Las Vegas and we have we're running customer workloads. We're live in those cities. Probably our most advanced city is Las Vegas.

There's a lot of activity going down there and we just announced that we brought another 26 municipalities up to what we call customer ready customer ready is our name for our Justin Time sort of production model where we build our data centers in the factory these you can think of them like our smallest status like the size of a UPS truck. It can fit in a parking space or larger ones like the size of a shipping container. Small ones are 20 kilowatts and a large ones are like 160 or 180 kilowatts.

And just in time means we the data center can leave the factory like in Phoenix get truck to the location. Maybe that takes one to five days. And once it's at the location we can stand it up in in three to six hours.

And the reason we can do that is because we spent the last five years pulling permits pouring foundations trenching fiber laterals bringing in the utilities everything needed up to the point of just laying in a data center. So our promise to customers is that any of the municipalities that are on our list that aren't live today we can have it live for you in 90 days typically and comparing contrast if you would if I wanted to replicate that same kind of capability using a public cloud service, what would be involved how much time and effort we require and well today you yeah. Yeah today today.

You can't exactly although the cloud providers are Target customers of ours and so as the cloud, you know today you think about Amazon, right? They started with two regions Us West and Us East and now they're deploying I think like to 12 to 16 local zones, which is sort of regional and then the next step is is like what we're doing with the Connecticut grid, which is like very very, you know, Neighborhood level and just to give you some sense of like the timing and our Network can deliver to the premises within our Geographic the footprint sub 100 microsecond latency, and that's microsecond not millisecond. I think everybody else in the internet will tell you we'll talk about millisecond latency.

You know, we're we can deliver three milliseconds or five milliseconds. This is a microseconds and we did that very intentionally to essentially operate it at land speeds so that you have this it's it's if the compute is actually on your premises, but it doesn't have to be And is the way we think about engaging with data changing and as that part of the reason that we're seeing this shift here because it seems to me at least that there's more of an effort going underway to process data and analyze it at the point where it is creative and consumed rather than having all these round trips out to the cloud and back and hoping the way it works, etc. Etc.

So are we kind of in some ways going back to the future but we're just finding a different way to skin this latency cat. Well, I mean in a way, I mean there's a lot of reasons for the edge data being one of them, right? I mean data, especially now with like a 4K cameras with being like the sensor of choice generates voluminous amounts of information that is very expensive to ship back to the core and it takes too much time to do real-time inferencing on that.

So definitely the fact that data is being created at the edge is a reason to have the compute at the edge. It's a lot cheaper to move a you know, a two gigabyte application to operate on 10 terabytes of data that is to do 10 terabytes dated to you know, a small application but you know the way to think about it I think is like like the, you know, we talk about the internet of things right the other things doesn't really exist today. Not not really not the way we imagine it do we have the internet and we have the thing and I think what Edge Computing really is is the merger of the digital and the physical world so that you can deliver the all the complexity and scalability and Performance of the cloud but right at the thing where it can it can operate directly up, you know with the physical world.

With us driving up, you know erotic lathe or analyzing real-time, you know video cameras to make sure everybody in the construction sites wearing a heart helmet. So do you think we'll see a whole new classes of applications that previously we wouldn't have attempted it or we just gonna take existing applications and move them a little bit closer to the point of consumption or wait. It's gonna be it's gonna be a mixture of both, you know, when we started this journey five years ago Mike, it's like everybody thought it's me autonomous cars and we know where that is now, right?

I mean eventually that is gonna be the application of choice, but it's it's not currently and where we found a man is really in in three places. One of them are Enterprises that have you know that are on their Cloud Journey have moved a bunch of workloads to Cloud but still have substantial workloads locked on the premises because they have to be there for performance and security reasons and and they need they want to get out of the data center business. They don't want to buy more equipment.

They typically don't have hyperkill class data centers. And so, you know upgrading to really beefy gpu-based service and stuff isn't really an option. And like I said, they don't they want to get out that work off the premises and they also want to consume things as a service right?

That's the Model and when you're buying the hardware and Licensing the software, it's it's just a headache that most Enterprises are trying to get away from and then you've got on the on the provider side these companies that either deliver traditional applications that can only run on premises and like Factory robotics or video surveillance. They now want to be in the business of offering this as a service because they believe they grow their their addressable Market by doing so and they create more convenience for the customer in a differentiation. Well, the competitors are trying to catch up so that's those are the existing applications.

And then of course, there are new applications. There's things that you know people that that have applied their Ingenuity to the cloud can now bring it to a new class of workloads that can only be delivered from The Edge, but they can use all the the cloud and Cloud native capabilities that they you know developed expertise and In that case if everything is eventually becoming a cloud do we need to use the term cloud computing anymore? We're just going to go back to something old-fashioned like yeah distributed Enterprise Computing and go from there.

Well, I think we're always gonna call it Cloud because maybe there'll be something else right? In fact, you know, we're a co-founder of an organization called the open grid Alliance and you know, we're we think the the cloud or all this compute should operate like a fungible fabric where you just have a workload and you pass it off and it figures out where to run and that's a vision. So maybe that'll be called something else.

But today the cloud is so important. I mean, you know, think of and I I wouldn't try to upset that that term but one thing I do think we'll stop doing this. We'll stop talking about Edge as if it's something different just be the internet.

As we go along where does 5G then fit in this whole equation is that guy live a whole new class of requirements or what's your sensor? Well, yeah. So so delivering 5G networks is a very very excellent application for infrastructure at the edge like the kinetic Grid.

2 split and a 5G which is sort of the open ran and that's where everything's running is software essentially, but you split up the pieces and so you think about like this old baseband unit to have all this complexity and technology and then Radiohead that all this now the piece these apart into mostly software not these dedicated appliances and with our model because we're sub 100 microseconds to the premises or to the the Tower with the with the with the small cell on it or to the macro cell. We can drive a 5G radio from the infrastructure with nothing but the radio out in the field, which is Really novel in use and changes the economics because now the economics of densifying a 5G Network first on shared infrastructure, right? So you're not paying for the whole thing yourself you're sharing it across multiple multiple delivery, you know vendors delivering different services and the fact that it's already built, right?

So time to Market it's already built at the density you need and that's I think it's going to be pretty game changing in terms of the capabilities of private networks, which often is private Wireless private 5G are very important to a lot of these new applications. I'll give you an example. So you go into a city like Las Vegas where we have a large kinetic grid.

There's maybe I don't know 7711s and you know 150 McDonald's I'm making this up but there's a lot of stores like that now it's imagine that McDonald's wants to roboticize their back a house, which they are planning to do and maybe seven eleven wants to do automated checkout. So out of me a check out requires like a camera A couple feet on the ceilings just like 75 cameras in a 7-Eleven in order to do automated checkout. Well, where where is that compute going to go?

Right? So it's bad idea to store to the store because like it's a Threat Vector and you know it you have to do truck rolls and go to every store. So you want to start pulling some of that into in a more into more Consolidated location so you could potentially serve, you know, 2711 from one location on the kinetic Grid.

It's supposed to just one. And reuse those those resources. The other thing with the 5G is in existing building.

It's actually costly to install cameras. The cabling is often often pretty complex and you know, like a school campus or shopping mall getting wired cameras up is a pain but if you drop in a 5G hub on a private Network, you can put 5G cameras in pretty easily just anywhere you want and so 5G is probably is becoming a conduit for these a lot of these services but it's not essential. I mean we take fiber to the premises and that's what actually happens.

That's where the data goes across but then there's of the if there's a if there's a 5G Network, it's typically a private 5G Network that then it's assisting some particular use case like controlling, you know, connecting all the machines to to the the kinetic radiator connecting all the cameras to, Connecticut. how small can Community grid get I mean you describe the container. So do I put a container in the neighborhood or can yeah, so we deploy now?

That's that's a great question. That's great question. So there's a lot of elements to that right?

There's there's permitting there's noise ordinances. There's you know, especially you're going into a neighborhood and so we had to count till for all those things and we have different sizes. So we have three basic models today.

We have a vem180 which is 180 kilowatts. That's a big one. That's the size of shipping container.

It's not a shipping container, but essentially like that size and we have a vm160 which is almost that big and then we have a them 20 and then 20 is Fitness parking space. And so they're relatively small they're relatively easy to for us to deploy their comparatively less expensive. They're easier to permit.

In fact, one of the most amazing things that you know, we can stand one of these up in three to six hours. So that is how you know, we get to to 32 markets 26 of our markets don't have data centers in them yet. Everything else.

They got the utility the permitting the foundation the fiber everything and we just dropped the data center on a truck and crane it onto the site and a few hours later. It's ready for customer equipment. What's the stop somebody with a forklift just coming along and picking the thing up and we but we bolt it to the I mean, it's hurricane rated.

It's bolted to the we have security cameras. Yeah, we these are designed to be run lights out with no Personnel on site and they're often in a in a compound that has you know, security fencing and things like that, you know think like the base of the cell tower or something like that. How do people get started what's the thing that you see them doing first?

And what's the thing you wish people wouldn't do and make you shake your head. That's a great question. I mean every every use case is different and it's it's you know, a lot of what we do today is we help our partners and customers figure out how to bring new services to Market.

So there's no one answer that I would say that the there's a kind of clever answer though that I'll give which is which is to change your way of thinking right. I mean, I think there's a lot of of providers of really interesting Services. There's still locked into this old way of thinking, you know, they think of like interconnection like well, it happens in Regional data centers, and there's only like 12 places in the US where you can do, you know meaningful interconnection on it in the internet.

Well, no thousands of places should be out at the edge be machine speed and so once you start thinking of these capabilities being available, it opens up, you know, the imagination to you know, wondrous use cases, but I think there's like this inertia that the industry the industry has which is to you know consider to use the old patterns as opposed to adopting the new patterns. All right, folks. I heard it here first there might be a very tasteful Artful boxing your neighborhood soon may look like artwork for all you know, but inside it will be a data center.

Hey, Matt. Thanks being on the show. Yeah, Mike appreciate a lot.

All right back to you guys in the studio.