Not Your Father’s Data Center - It Always Comes Down to Power

Episode Date: May 25, 2021

It is hard to believe, but there is a significant amount of megawatts that goes unused in global data centers. On this episode of the Not Your Father's Data Center Podcast, Host Raymond Hawkins of Compass Data Centers talked with Dean Nelson, the CEO of Virtual Power Systems. Nelson talked about his work at Uber and Sun Systems. This led down the path of Nelson's work at Virtual Power Systems, where they "believe that digital infrastructure is the foundation for an equitable and inclusive world, where every person on the planet participates in the digital economy." In acting upon this vision, VPS unlocks stranded power in data centers. What's staggering to Nelson – and why he thinks this is such important work – is that, through research, VPS found 35,000 megawatts of capacity globally in the data center industry. They checked this data through multiple vectors. Of the capacity, 10,000 megawatts are stranded. This leads to a host of issues, one of which is increasing the carbon footprint unnecessarily. "How much embedded carbon do we add every time we perpetuate that problem and build another data center that's 50% utilized?," Nelson asked. The goal through technology at Virtual Power Systems is to utilize this space. Data centers build such an extensive infrastructure with low utilization to make headroom and safety create into the center. They want to make sure they have the buffer room when things don't go right in the system. "It's the cumulative effect of everyone's buffer," Hawkins said, referring to the sheer amount of space unused.

Welcome to Not Your Father's Data Center podcast, brought to you by Compass Data Centers. We build for what's next. Now, here's your host, Raymond Hawkins at Compass Data Centers. We are recording today, Thursday, April 22nd, as the world continues to fight against the global pandemic, month, I would say 16 now, I think, 17, where we continue to climb out. Today, we are joined by the CEO of Virtual Power Systems, Dean Nelson. Dean, thanks for joining me today. Yeah, great to be here. We are excited to talk about all kinds of things with Dean because Dean is infinitely smarter than me and he will say things that I don't even understand, but I hope that our listeners will track along.

You can, once again, answer our trivia questions by either emailing me at rhawkins at compassdatacenters.com or you can tweet us at CompassDCS. So like always, we're going to open up with three trivia questions. Then we're going to let Dean run the show. In honor of Dean's time at Uber, we're going to ask Uber-related questions. So Dean doesn't get to answer.

He is not eligible for the $500 Amazon gift card. But the questions, number one, yeah, I know, heartbreaking. I know, Dean. What is the longest ever recorded ride for Uber? So that's trivia one, number one. How many trips a day at its peak does Uber handle? And then what year was Uber founded? And like always, we'll save trivia question four for the end.

You got to email us all four questions correctly and you enter it in for the drawing. All right, Dean, that's some fun Uber stuff. Dean, if you don't mind, we've got folks that listen to us in Asia and in North America and in Europe. And I think a lot of folks know who you are through either your time at Uber or at Iron Masons, but a lot don't. So can you give us the Dean Nelson background, home, grew up, business, all of that. And then we'll kind of get into talking about what I loved your

phrase yesterday, the data tsunami. So tell us a little about you. Yeah. So I was born in Minnesota and moved to Colorado when I was six, moved out to Arizona to go to school and then hit the California coast. And I've been here for 31 years ever since. So yeah, I'm old. But you got out there and liked it. Oh, yeah. This is a great place in the world. But my career, I've been doing this for about 31 years now. So I did 17 years at Sun Microsystems. I started on my 21st birthday in 1989, which was actually really cool. Scott McNeely.

All right. Yeah, I've done podcast interviews with Scott at our iMasons events and things. Andy Bechtolsheim, you know, Vono Kosla. Yeah. Bill Joy. Like just, you know, all those folks. It's a great thing. And I always say that I went to the University of Sun, by the way, because I learned business and just strategy and all the other things.

Technology at Sun, it was such a great foundation. But I spent 17 years there. What a great time, too, to be there. Yeah, through the 90s. I mean, what a great time to be at Sun. Oh, the growth was incredible. It's kind of like the growth now. Yeah. And they were just such an innovative company. I just love the culture and what they did. I mean, they dominated the internet. If you think about it, they put the dot and dot com.

I was going to say, remember the dot and dot com ad. Right. That's right. We're the dot and dot com. That's right. Yeah. And then when the dot bomb happened, actually, ironically, I left and went to a startup company in 2000 and was in a bubble of my own there for three years, which is incredible. But then, you know, the how would I say this? There were some factors that basically killed that startup. And that was my first experience with a startup. That was really I learned a lot during that.

I had a great time, worked way too much, but loved every second of it. I went back to Sun, spent another five years there. And then I left when Oracle took over. And I went and joined eBay, spent seven years at eBay doing global foundation services. So data center, hardware, network, supply chain, running the budget, strategy, forecasting, all those elements to say really metal is a service. And then I went and left and went to, I took a six-month sabbatical between eBay and before I went to Uber, went and looked at colleges with my daughter and all that kind of stuff. It was great. I also started Infrastructure Masons at that point on April 2nd, 2016. And that was really to get my community back together. But then I went and joined Uber and did that rocket ship for three years, which is incredible.

I left in 2019 on my 51st birthday. So 30 years exactly in the industry, which is really cool. And I wanted to move into something completely different. So I started doing advisory work and joining boards and those types of things. One of them was virtual power systems. And during that initial period, they asked if I would step in as interim CEO. I said no five times, by the way, because I'd done 30 years of operations. But then I said I would. I stepped in and started doing this and found that I actually loved it. So, you know, I took the permanent CEO role in August of last year.

We did our first raise. I'd never raised money before, you know, so totally different world. Just such an important thing that we're working on right now and such a massive market opportunity. I just love being in the middle of all of that. And of course, still doing advisory work for some Fortune 100 companies and other large private equity, as well as iMasons. So, that's a long-winded answer to your question. That's a great background. Super. I'm doing the math as I'm sitting here listening to you. 67? What year were you born? That's me. I'm 67. 68. 68. Okay. So yeah, I was born July of 67. So it sounds like we're very, very close there.

As you went through the history, I stumbled into the technology business, complete accident in 1986. And I hate to do the math and realize how long I've been working. But yeah, I walked into the Mac lab at my university and said, you know, hey, these are interesting and fun to play with and ended up working in the Mac lab and becoming an employee of the university. And think about it, the Lisa came out in 84. And so we were still very early in the personal computing world. And at that time, at that time of the university, the computer lab was the mainframe. That was the computer lab. This little Mac lab was like a play thing. And that was the beginning

of my experience in technology. So yeah, very, very similar life trajectory. And, you know, I think when you and I both refer to Y2K, it means something different to us than people that are running businesses today. They call that the biggest non-event in history, by the way. Yeah. Yeah. Yeah. I was, I was sitting in history, by the way. Yeah, yeah, yeah. I was sitting in a data center at midnight, that's for sure. Yeah, ditto. That was, yep. But nothing happened.

You touched on something important there, which was happened into the industry. So exact same thing with me. I just interviewed with different people. I went and got an associate's degree in electronics from DeVry University in Phoenix. I spent two years getting that one, and then they hired half my graduating class. And I moved to California. I had no idea what Silicon Valley was. I just learned everything right there. You know, that University of Sun was real. Well, I'm with you. Getting to work at Sun and getting to see, I mean, what Scott did in that whole business was just fascinating how they, you know, not only at the

early days, but also the early days of the internet, how they seeded so many companies with compute power. I mean, we could get way sidetracked in that. We're going to have to end up doing three or four episodes, I'm convinced. Let's stick to the focus. Let's stick to virtual power systems. Appreciate your background. as you talk about why, and I like you said, it's important work and important, not just because let's help businesses optimize how they manage power, but let's be good stewards of the resources here on our planet. So talk to us a little bit about what virtual power systems does.

Yeah. So we're a startup company that basically is virtualizing data center power infrastructure. The punchline is pretty simple. We unlock stranded power in data centers, but we do that through a hardware and software combination. But let me back up to the stats. What's staggering to me and why I think this is such important work is it ties to iMasons and our sustainability vision of every click improves the future. And the reason it does is that we've done a research and found that there's a baseline today of about 35,000 megawatts of capacity that's built in the data center industry globally today. And that came from multiple vectors that we validated, everything from UPS sales to power draw, those things. And we've correlated all those back into this 35,000 megawatts.

But the challenge is that I believe that at least 10,000 megawatts of that is stranded. 10,000 megawatts. And so if you think about it, we've got 10 gigawatts of capacity out there. But now there's, there are very good reasons of why it's stranded. And that goes into the technology side. It really is about the stack. You know, in my, in all my jobs, we've been trying to drive efficiency and sustainability. I mean is about the stack. You know, in all my jobs, we've been trying to drive efficiency and sustainability. I mean, from the beginning, there was a guy named David Douglas at Sun Microsystems where he really went back and said, this is about economics and ecology. You have to balance both. There's not a choice between them. And he was absolutely right. And it just took

focus for people to go back and say, I can make a sustainable choice in how I drive, I build, I operate data centers and infrastructure. And so when I look at that one, tying back to this strategy, if there's 10,000 megawatts worth of stranded capacity, how much embedded carbon do we add every time we perpetuate that problem, build another data center that's 50% utilized? This isn't people just being bad about, you know, I just want to be inefficient. It's literally the customer behavior. When I hear the term stranded power as a data center guy, I think of somebody who I've built them a data hall and we put four megs of capacity in there and they filled the data hall with racks and their racks are all done.

They can't, they don't have any more physical space, but they're only drawing down three megs of power. So that megawatt is stranded. I built the infrastructure, but they can't use it because they don't have any, they can't put any more physical compute in the room to utilize that other megawatt. I think of that as a traditional stranded. I think you're talking about that, but also something else. Is that true? Yeah, it's a little bit, let me fine tune it. And I'll give you an example. So at my last two jobs, we would build out these standardized zones for cloud deployments.

So we had an on-prem, right, cloud instantiation. And so we would say, for example, at Uber, we'd have 576 cabinets. That was a zone. There's 480 cabinets that were for the compute, the storage elements, the standard components. Then we had 32 racks that were actually the network. And then we had 64 racks that were going to be for overflow and flex and lab and test environments and that kind of thing. Right.

So when we roll that in, we had really tightly coupled hardware to the shared platform that the developers would be utilizing. And the whole point is the shared platform was shared. And so we would be able to drive the utilization up, et cetera, by matching that. Then, of course, we optimize with the hardware and then we matched to the physical data center power footprint. So when we'd roll those out, the concept here is that we'd be able to get high utilization.

But the reality was this. We still have multiple locations and we have this replication factor, which means I have to have a certain amount of stuff in multiple places in case I've got a region fault where I have more than two zones that fail in a region, I have to fail over to the other region. And so the problem is that everyone would buffer the buffer

and the buffer and the buffer again. And that's what leads to it. So low utilization is usually mismatched because of just the applications and the type of hardware, but also because they have disaster recovery and all this headroom, this safety built into it. And the problem is that we keep perpetuating this problem again, meaning that-

Yeah, Dean, no one gets fired for buying too much. People get fired for when it goes bump in the night, it doesn't work. That's what people lose their jobs over. So they buy all this buffer room. They're buying buffer room as, frankly, job protection, which I understand the motivation, but it may not be the most cost-effective

or the best stewardly decision. Right, and this sounds very familiar to trends that happened 10 years ago which was that's my server i'm unique that's my network that's my story i can't no no no no i can't share with anybody else i'll have noisy neighbors you're going to don't put another application on here i don't want it bumping in with my application because then when something goes wrong i don't know who to blame that was why people would argue against virtualizing a server or virtualizing the network right and so when you look at it today, it's the de facto standard.

I mean, you wouldn't go build dedicated hardware for all of your different applications and things. It just that's what cloud's based on. That's what all of the efficient, you know, even on-prem data centers are doing. It's back to your shared resource that you did in your in your design, right? And we're just doing that everywhere now across the network and across the infrastructure, the computer infrastructure. Absolutely. So the critical part to really look here is that it's not the data center engineers or operators behaving badly. It is that they have things pushed to them that they have no other choice but to manage. In other words, when they say this hardware is going to

draw 15 kilowatts and it draws 5.3, they still have to be able to allocate more in case something draws that kind of power. But here's again the reality. We had a full region failover, right, that I've experienced multiple times. And people think that because of disaster recovery, well, if I've got all these load in one place and it fails over to the other one, I'm going to have a doubling of the demand on the other side. So I have to have all that headroom. Right? Right. We had a failover, peak volumes, and we had less than 10% increase in actual draw in the other region with the whole world failing over.

Wow. Think about that for a second. So it just tells you an underutilization. Yeah, that's right. It shows the underutilization, yes. Yep. So, whether it was the data center to the hardware to the shared platform to the applications, those buffers all the way up. So, low utilization is a big one, but so is redundancy and buffers. So, when you add all that up, that's where this 10,000 megawatts is coming from. Yeah, the cumulative effect of everyone's buffer. Right.

The cumulative effect of everybody's safety net. That's right. Yeah, that's a great... I love that analogy that when you brought that over, even though... And let's just make the number simple. Hey, we've got a megawatt load here and a megawatt load here. This one goes down. We're going to move it over.

Well, wait a minute. It was a megawatt on faceplate. It wasn't a megawatt on draw. Once all the, once all the headroom got taken out and it was just actual load, it came over and it was 300 kilowatts. That's really, I mean, to put numbers on it, I'm just making the numbers up, but that's what you're saying. Yeah. And so your megawatt over here that also only had 300 kilowatts of draw now goes up to 600 and you still got headroom. And it's that collection of headroom on both sides. Wow, okay.

Yeah, but also, that 300 and 300 don't go to 600. That's right, there's redundancy in there as well. That's right, that's right. So the whole point of what we're doing here is that applying software-defined methodologies to a data center seems weird. Everyone looks at it as it's a physical environment, like I have these dedicated elements. But you have stranded pools of power everywhere in the data center. So the way we

approached it is that you have the ability to have distributed capacity. So we have the thing called power bursting. That is actually installed right at the PDU or RPP. So we inject current in parallel. We add power where you've got the stranded capacity. I got you. That's one element. Then we have another thing called phase balancer. So anybody who's run data centers has had to deal with the rebalancing of power because of customer decisions or single-phased hardware or low utilization changes. You physically are now moving them from one panel to another. So what our product does here is it takes this imbalance

and corrects it at the PDU. So upstream to the UPS, perfect balance all the time, 24-7. Oh, wow. This phase, and by the way, every data center in the world has phase imbalance problems. It just comes down to percentages. So I look and by the way, every data center in the world has phase imbalance problems. It just comes down to percentages. So I look at that going just now the sheer nature of people making decisions on, I have to have this pair of PDUs with these racks over here with this nameplate expectation.

Even if I derate it down to this amount, I'm going to have this kind of buffer in it. So they do buffers and buffers. Then you've got the low utilization on the compute. And then you have the redundancy. Again, you never get past 50% in aers of buffers. Then you've got the low utilization on the compute. And then you have the redundancy. Again, you never get past 50% in a pair of PDUs. And most likely it's 20% to 30%. So you have all this stranded capacity at that pair of PDUs. Well, how do you get to it?

Well, you inject current and you balance phases. And you reclaim 30% to 40% safely without compromising the SLA. Because this also improves the SLA. Because you have distributed capacity. It's not a UPS. It's meant to augment because it's for those ride through moments. When I have a failure on A and things fail over to that one, if I had imbalance and I'm at 93% utilization on those circuits, I'm going to trip the other side. If I balance it, I'm at 84%. And then I've got surge capability within power bursting. So in the end of it, you're actually increasing your SLA.

So our software orchestrates all those elements. So Dean, you said something at the very beginning as you were describing the bursting and the balancing that I thought was great. You said, you know, Raymond, people ask, hey, we've got physical elements in there. How can we virtualize them? Well, Dean, we got asked that

when we were talking about virtualizing components inside a server. We got asked, oh, no, no, no, you can't give my resource. You can't virtualize my memory. You can't virtualize my storage. You can't virtualize my network. We've lived through all those iterations. Hey, a network is a physical asset. But now people, we had to live through getting people comfortable that we could virtualize the bandwidth inside of a network. We didn't have to give you a dedicated port that nobody else touched that port, right? I mean, so it's interesting because I'm in total agreement with you that you have to first get people comfortable that, hey, your physical asset can be shared without impacting your SLA.

That got lived through memory, disk, processor, and network. It's the exact same conversation. And if you think about the parallels to that, you had to win the hearts and minds of the engineers because they were the ones going, my job's on the line, not yours. Don't take my stuff. Don't take my assets. I need to hold them. Yeah. I'm special. And so, the thing is that you apply that same methodology back over to the power side. Think about, like you said, nobody gets fired for over-provisioning or for being conservative. That's right, that's right.

But flip this around now. So here you've got the, okay, this is the product, the engineering challenge, the inefficiencies that happen in a data center because of behaviors of other people. And by the way, for co-location companies, it's all because of their customers. And I was that customer.

And they're trying to take care of that customer. That's right. Right, because in the end of it, the SLAs that they've got is penalties if they don't meet those SLAs. So there's a very, very real business case behind why they do what they do. But if you start to apply that methodology to it, all of a sudden flip around to the business side. the capacity I built in a data center at $8 to $10 million a megawatt. The yield, the return on that investment itself, I have half of that money sitting on the table. So imagine if you're

able to go say, I could actually drive the utilization up overall as a colo business safely without compromising my SLAs. Here's the other thing that's really interesting too, that I don't think people have quite grasped, a few have, but others in the industry haven't yet. If you are able to do that, your cost per kilowatt can go down in your RFP responses. Yet your margin can be maintained or increase. And your performance can be maintained or increased.

So suddenly you have a competitive advantage. So that's on the colo side. When you look at it, it's like, how do I go back and give low prices to my customers, but still make money? Because there's a race to zero right now, especially in the hyperscale side. Our industry is getting commoditized. There is no question. That's absolutely right.

Yet it's growing like never before. It's so big, right? And so when you got markets that are really getting hit on that cost per kilowatt, how do you stay competitive? You must rethink how you're doing these models because the fixed element of doing it this way, dedicated power and things are just, it's going to impact the business. Yeah. And Dean, I like that, you know, when we talked about virtualizing inside the compute world, you nailed it. We had to convince the engineering side of the house. Those guys had to get comfortable that they had to lose the I'm special badge and go,

no, I get how I can have the cycles I need virtualized. They had to buy off before anybody else would sign off. And we're going to have to live through the same thing on the, on the, at the data center power level, people are going to have to the smart guys who say, yep, that's resilient enough, that's available enough, that's enough capacity are going to have to understand and get what we're doing here. Yeah. And when they do, I'm with you that there's, I like the fact that, hey, we can do it at a cost effective manner, and still be able to have a margin because the end of the

day, the industry needs us data center providers to be profitable so that we're here when next year when they grow again, and the year after that when they grow again, because I teased with the data tsunami comment, because that's coming, right? The data tsunami is coming. And so I really, really, really love the tutorial around what virtual power systems is doing. We may just do two or three shows just on that on bursting and on how to keeping things in phase. I think that'd be fascinating. Let's switch gears and talk about the data tsunami, because at the end of the day, we're talking about being good stewards with that power. We're talking about how do we go get that 10,000 megawatts and utilize it better.

And that's what VPS does. Let's talk about why it's so important because of what's coming. Yeah. So IDC predicted, a lot of people have seen these things, you know, that we would have 44 zettabytes worth of data generated every year by 2020. And so they just updated their report. And this is all pandemic driven, by the way, if you think about it. It was 64.2 zettabytes generated last year. Okay. So it wasn't a 10x increase because they had 4.4 zettabytes back in 2014 or 2015, whatever it works out to be in that, in that timeframe, right?

10X growth. So it was actually a 16X growth now to 64.2. Okay. They had predicted with that 10X growth, 175 zettabytes of data generated every year by 2025. Okay, which is huge to begin with. If you do the math itself, it's massive amounts of data. But now if you extrapolate and take that same 16X, we're going to be at 255 zettabytes by 2025. Now, that's assuming that we're doing the same things we're doing today. Now, people have said that the pandemic was kind of a one-time event for saying this. They said, Raymond, we've seen three years of IT transformation in three months.

And I completely agreed with it then. And it's continued for another year. But to your point, we're not going back to that. In other words, we're not going to undo that transformation, right? We're not going to stop doing Zoom calls. We're not going to stop doing work from home. We're not going to stop doing telecommute. To your point, that's the new baseline.

From how much data and how we interact with the data and how we access the data. We're not unwinding that three years of transformation. No, absolutely not. And this is the thing. That represented half the world's population. There's at least another 2 billion people that are coming online in the next couple years across APAC, right? LATAM, Africa. Alone. Yeah. Just new users. Forget any other transformational effect. Just new users are coming. And their devices with a device in hand or like us in America, multiple devices in hand. Yeah.

And this is where I think people also get a little mixed up is that um you know the internet was built for people to do things right but there's only eight billion only eight billion the new internet what's what's going on right now what's happening with 5g what edge is going to be what's happening with with the edge deployments and also with quantum computing because that's just going to explode processing this is about machines talking to machines. This is not about humans downloading movies faster. Everybody uses that analogy. That's right. And it's so narrowly focused because today we think of connected, right, you're talking about zettabytes. It's really a function of how many devices are interconnected.

And we think of devices interconnected as my iPad, your laptop, my phone, your car, my Nest thermostat. That's what we think of connected devices. But that Nest is the beginning of devices talking to devices. And that growth, you and I may have five or six personal devices, but that growth of devices talking to each other is where we're going to see this exponential growth. Yeah. And there's a couple of factors that are rolling into this. So if you think about, there's 8 billion people and half of those people are on the internet right now. So great, that's going to drive up. We'll have more use, but they're predicting right now, there's going to be 125 billion things, the internet of things connected by 2030.

But so just think of the ratio. You've got 8 billion of people connected through devices. Then you've got all these actual things. This is smart cities and all the other initiatives. This is also smart factories, smart farming, smart everything. And your refrigerator and your television talking to you. Those are everything. Yeah. So just every household will have hundreds of devices. But think about what's behind that. They're saying there's going to be over a trillion sensors of those things communicating. So again, the machines are talking to machines. Why are they talking to each other? To enhance the human experience and increase engagement. So it's not about the movie download. It's literally about how are you going to change your behavior? So let me give an example of this.

You know, when the, I guess I get a lot of people debating about edge. Oh, there's nothing there. We could do it with the cloud. We could do that, right? It's frustrating in the conversation because I don't think people see it yet. The reason that the state of tsunami is really coming is that those smart cities and those initiatives are going to be generating massive amounts of data, but not all that data needs to go back to the core. Hear, hear. About 70% of it will be dropped.

And the analogy I use, when I was over at Uber, we built up the depots for the autonomous car data ingestion. So we had to go back and orchestrate to be able to say that, you know, X amount of hours a day, they come in, they land. But the cars were generating so much data that they couldn't do it over their cell connection. Yeah. The physical download had to be connected. Right. So we'd have all these cars come in, and then we had surges.

And they would now, you know, we do 100 megabit bursts, right? Or 100 gigabit bursts out of these cars. And so four terabytes a day each, etc. They come in, dump data, go out again. But all that data when it comes in, you don't need all that data. If you think about an autonomous car, there's like eight cameras on it and they're recording everything. But the same car is driving down the same road doing the same thing. Unless they see an anomaly or a delta, meaning I could not interpret that thing, that person, that movement, that interaction. All that other data, you keep a copy of it, not 150 copies from those cars that are going 24-7. So there's a huge compression that's going to go down. But where is it going to be done? All at the edge. Right. Right. Right. Because the cars won't have enough power in them to actually do that de-duping.

Right. And then when it gets into the edge deployment where they're going to have this on compute, that's where this thing's going to say, drop, drop, drop, drop, drop, drop, drop, send what I care about. Because if all that data had to go back. Send change. Yeah. Send the anomaly.

That's all that's going to have to get transported because otherwise we'd crush the network with that. I mean, you talked about the car. I think you said four terabytes. I mean, people can't comprehend how long it takes to download four terabytes. And to try to do it, I mean, I'm got it. 5G is going to be awesome. But trying to do it wirelessly, it's just, it's an, and that's just one car on one day's drive. I mean, the amount of data is astronomical. And that's the part where the other element that is really going to blow this open, everybody talks about 5G and all these advancements, but when you break it down and you look at what 5G is going to bring, it's going to be 10 times faster from a latency standpoint. 100 times more bandwidth, I can basically go to New York 10 times faster, and I could do it in 100 lanes wide. Wow, right? Yeah, that also is 100 times more concurrently connected devices. Think about that. This is why all these things out there. Now it's not again, you downloading the movie faster. Great. It is

about your fully immersive experience that's going to happen at the edge. And the analogy I give in this one is that just think back to when the app store started. Did people believe there was going to be billions of apps and everybody was going to be engaging in that? No. Why did it happen? Why did it work? Because they had an open platform that allowed all of the developers to go back and do something. So when 5G opens up those floodgates, what are they going to do at the edge? They're going to come up with millions of experiences that are going to engage everybody, machines, all that. Here's the data tsunami. Yeah. You go back to the beginning of the app store. I remember when people joked, oh, yeah, there's an app for that, right?

It was sort of comical of what new ways of interacting with our world, which is what you're saying. You know, there's a different way if you digitize this interaction. I mean, I'm going to give a silly one. Waze is a perfect example. There in my pickup truck, there are two folded maps. There's a map of the state of Texas and there's a map of the United States. And my kids literally don't know what that's for. They pick it up and they fold it all out. And they're like, dad, what is this? And I'm like, that's a map guys.

And they're like, well, what did you use this for? And I'm like, what do you mean? I mean, the concept of they're like, it's so you could figure out how to go somewhere. And they're like, dad, why would you not do that on your phone? And think about it, Dean, that's just been in 12 years, right? 12 years ago, 15 years ago, there wasn't an app store, right? There wasn't a smartphone. And that's a simple real world example. You don't use a map anymore. You use Waze. I'm still driving somewhere. I still need directions, but the way I interact with it, and what my map told me was this is where the road went through Jackson, Mississippi to get me to Atlanta. Now what Waze tells me is there's an accident three miles ahead.

There's police seven miles ahead. Not that I ever speed. There is a traffic jam. The road construction is slowing you down. It's changed the way I interact with what used to be a hard to fold up map. Right. And that's a simple real world example of everything you're saying. Right.

When you unleash the thought process of 8 billion people and go, how can I interact with my world differently? And oh, by the way, like you said it, I now have 100 lanes and I'm going 10 times the speed. Instead of 60 miles an hour, I'm now going 600. And instead of two cars, I've now got 100 cars. Things are going to change. There's my prediction. Things are going to change. You know, Raymond, I think you might be right. Yeah, yeah. I could be onto something there. Well, I mean, I've made some bold predictions before, Dean. So, I mean, you know, I do joke occasionally when people ask about our business, I'm like, you know, we're pretty bullish on the internet. We think this thing's going to stick

around. So, so I actually want to touch on something that you said there too, because this, the, the edge definition, you know, we talked about people saying movies and they don't need to have this, but the, the, everyone says that the edge is going to be about a ton, fully autonomous vehicles and flying cars and all those really cool whiz bang things that are coming. They will be, but not right now. And so people are saying, well, I don't really need it yet. Well, what you do need, so take the autonomous car example. Now take the autonomy out of it. Just stay smart vehicles. Like I have a Tesla, right? And so tesla has a certain amount of power in it

it has a certain amount of processing it has a certain amount of sensors lidar or all the different elements within it and then it can it can calculate and do things but it's autonomous meaning it makes its own decisions right and you're still interacting with a human but the edge has got real use cases today that are requiring compute to be very close. The number one is public safety. So cities and smart city initiatives are rolling out edge because whether it's people walking or cars or anything with emergency vehicles that have to get through something, it's about traffic management. Edge is going to give that real time thing. You gave an example of Waze, right? And so when you suddenly have the ability for all of these smart vehicles and smart intersections to be able to now see who's

walking, see who's driving, and mesh networks that allow them to communicate with each other instead of going 150 milliseconds round trip to the cloud, that all of a sudden you have this ability for them to orchestrate. And natural disasters are one of the big ones that come back up. If we have some type of tsunami or not tsunami, if we have some type of a hurricane or other things like Katrina and those things,

we lost connectivity, we lost power, we lost all those. Edge is gonna provide the ability to orchestrate, to keep those things. We lost connectivity. We lost power. We lost all those. Edge is going to provide the ability to orchestrate, to keep those things up. Drones will be able to go out immediately from those things to find out emergency situations where people can't. The orchestration of all that stuff in emergency situations is really critical. Then there's data exchanges between all these different departments at city governments. And look, the point is, there are hundreds of billions of dollars of investment coming into cities, and that is going towards edge computing, connectivity, and the ability to orchestrate with these actual divisions. Then you've got all the commercial side of it. If there is compute on the corner, do you think that that restaurant, that coffee chain, that other ones,

and they could use that one? If it was available to them, what would they do with it? Kind of like the app store. Well, yeah, not only what would they do with it, but what do they not need now on-prem anymore, right? They now go, hey, I can get this as a service. And that's the thing I think that when we,

when this infrastructure gets built out and when people truly, just like you said, we got to convince the engineers. When people regularly go, okay, you know what I can get, just like when I turn my light switch on, I always know the electricity comes on here in North America most of the time. I know as a Californian, you don't always have that, but says the Texan with a seven day blackout. Wait a minute, wait a minute. Yeah, I was going to say. Yeah, that's exactly right. But when people trust that, hey, I can get the compute for my business on the corner from that locally provided device out on the street corner, and I can build a service, a reliable, predictable service, and I can build around it, that's coming, right? That ability to go, I don't need my own cash register.

I can do a thing there that allows me to interact with my customer. And I think about the ways to interact with my customer. I've got a sensor up there and my guy's got an app and he says that he comes to my coffee shop on a regular basis. Well, I'm going to know when he starts walking towards me. When he's three blocks out, I'm going to crank up his, you know, I don't drink coffee, but his mocha latte thingy, right? The ideas are limitless. And I like your point about safety and traffic flow. It's, you know, you made a comment earlier, every click improves the future. We like to say at Compass, we make lives better and truly do, right? If you could manage the fire engines to a fire faster because of managing, you've saved lives, you've saved property, you've saved people's

livelihoods. If you could get somebody, you know, if there's a mass tragedy and you could get people routed to the right hospitals where there's emergency room staff and the appropriate kinds of care, these kinds of decisions where now are getting done on radio dispatches can be done not just by recognizing the traffic patterns, but by managing those traffic patterns and sending a message to that ambulance, turn right here and go that way. That's the shortest route to the emergency room with the right facility. That's the kind of thing coming. All these pieces are going to tie together and they're going to tie together at the edge. And it's because about local decisions to be able to be made immediately. Now, one thing I want to clarify. So this goes back to the app store again. If the app store didn't exist, we wouldn't be able to do the things we're doing now. If the edge rolls back out and those companies that have all these retail locations out there suddenly have an option to use it, it's not about them getting something cheaper. The cheaper is going to be there. But also, they suddenly have the ability to say, oh, I have that one.

What else could I do? I talk to these companies a lot and they say, I can't justify edge computing in my store because I don't really have the use cases for it. We'll flip it around and say, you have access to all this capacity, literally a hundred feet from you. What could you use it for? Like, oh, well, I could do this with inventory management. I could have all my associates actually interacting with the customers longer or faster and more than they would have doing the back office things. Interesting.

So it opens up and it allows them to do things. Because just like I said with 5G, when you suddenly have compute and that speed, right, and performance at the edge, those developers are going to create things. I always use this example like- Things we've never thought of. Right.

Pokemon Go went overnight massive. That was one application. Well, suddenly when it's 10 times faster and 100 times more bandwidth, what do you think they're going to do? Massively immersive experiences that are going to drive engagement from people everywhere

and drive more and more things and more and more sensors to consume more things. Hear, hear. Well, Dean, I want to be sensitive. I know you've got a clock we got to manage. Thank you so much. I think we could do two or three more hours and we'd love to consume more things. Hear, hear. Well, Dean, I want to be sensitive. I know you've got a clock we got to manage. Thank you so much. I think we could do two or three more hours.

We'd love to have you back. Love hearing about what Virtual Power Systems is doing. We want to talk about iMasons the next time we can get you. This data tsunami is a thing. The edge is a thing. There's so much knowledge between your ears. We'd love to have you again. We're going to sneak in one more Uber-related,

in honor of your time there, trivia question. Question one, longest ever Uber ride. Question two, at their peak average rides a day. Question three, the year they were founded. Question four, who was that founder? Dean Nelson, CEO of Virtual Power Systems. Thanks so much for joining us. And we look forward to having you again in the future. We're not making you commit to it, but we'd love to have you back. Thanks, Dean. Thanks, Raymond. Take care.