This episode of Over the Edge features an interview between Matt Trifiro and Mike Conlow, Director of Network Strategy at Cloudflare. Mike started his career in political technology and was the deputy CTO for Obama for America, later joining Cloudflare in 2021. Matt and Mike discuss the future of the internet, closing the digital divide, and the work that Mike is doing at Cloudlflare to build a faster, more resilient internet.
This episode of Over the Edge features an interview between Matt Trifiro and Mike Conlow, Director of Network Strategy at Cloudflare. Mike started his career in political technology and was the deputy CTO for Obama for America, later joining Cloudflare in 2021. Matt and Mike discuss the future of the internet, closing the digital divide, and the work that Mike is doing at Cloudlflare to build a faster, more resilient internet. They also get into why there is a lack of wired broadband competition and network security solutions.
“Once you have a very complete edge network globally, all of a sudden running that logic about what is attack traffic and what is not, we can deliver that in a very performant way from our edge network. And we don't need to have a whole bunch of very expensive, very, very fancy on-premise boxes doing that work. We can do it with computing power on our edge.”
“We should make part of the story about 5G and about edge computing just the whole internet is going to get a lot faster to the point where it should feel mostly instantaneous.”
“12 million people in the United States don't have access to adequate internet, meaning that they don't have a wire outside of their house or a wireless way to connect to really good internet.”
---------
(02:00) How Mike got started in tech
(04:13) Mike’s current role at Cloudflare
(05:04) The range of services at Cloudflare
(07:36) Defining SASE
(08:35) The Cloudflare network
(16:12) How Mike thinks about edge
(25:00) The future of faster internet
(30:00) The digital divide
(34:00) Why is there a lack of wired broadband competition?
(39:00) What should we do to improve the internet?
(44:43) Magic Transit and denial of service protection
(47:41) What Mike is excited about in the future
--------
Over the Edge is brought to you by Dell Technologies to unlock the potential of your infrastructure with edge solutions. From hardware and software to data and operations, across your entire multi-cloud environment, we’re here to help you simplify your edge so you can generate more value. Learn more by visiting DellTechnologies.com/SimplifyYourEdge for more information or click on the link in the show notes.
--------
Follow Matt on LinkedIn
Connect with Mike on LinkedIn
Connect with Mike on Twitter
Narrator 1: [00:00:00] Hello and welcome to over the edge. This episode features an interview between Matt Trufiro and Mike Conlow, director of network strategy at CloudFlare. Mike started his career in political technology, serving as the deputy CTO for Obama for America and transitioned into the network space in 2021. He and Matt discussed the future of the internet, closing the digital divide and Mike's approach to the edge.
But before we get into it, here's a brief word from our sponsors.
Narrator 2: Over the Edge is brought to you by Dell Technologies to unlock the potential of your infrastructure with Edge solutions. From hardware and software to data and operations, across your entire multi cloud environment, we're here to help you simplify your Edge so that you can generate more value.
Learn more by visiting delltechnologies. com slash simplify your Edge. For more information or click on the link in the show notes.
Matt Trifiro: Two years ago, when I started the Over the Edge [00:01:00] podcast, it was all about edge computing. That's all anybody could talk about. But since then, I've realized the edge is part of a much larger revolution.
That's why I'm pretty proud to be one of the founding leaders of a non profit organization called the Open Grid Alliance, or OGA. The OGA is all about incorporating the best of edge technologies across the entire spectrum of connectivity. From the centralized data center to the end user devices, the open grid will span the globe and it will improve the performance and economics of new services like private five G and smart retail.
If you want to be part of the open grid movement, I suggest you start@opengridalliance.org where you can download the original open grid manifesto and learn about the organization's recent projects in activities, including the launch of its first innovation zone in Las Vegas, Nevada.
Narrator 1: And now please enjoy this interview between Matt Trufiro and Mike Conlow, Director of Network Strategy at Cloudflare.
Matt Trifiro: Hey Mike, how you doing today?
Mike Conlow: I'm doing great. How are you, Matt?
Matt Trifiro: I'm doing awesome. One of the questions I love to ask people, and I'm super curious about yours, is how did you get started in [00:02:00] technology? I
Mike Conlow: think I have a bit of a kind of uncommon story, both for my start and for where I am now, but I started doing political campaigns and I was a volunteer and I was always kind of the best computer guy who was around.
And so I started doing more and more data stuff for political campaigns and turn that into a job doing technology and turn that into a career in technology. So you weren't
Matt Trifiro: a technology person before you started doing stuff for these political campaigns? I
Mike Conlow: mean, I've always been a technology person since I was a young kid and had a Philadelphia Flyers webpage in the early 90s, but I am not a computer scientist by training.
I'm an economics person. I worked for the Federal Reserve. It was my first job out of
Matt Trifiro: college. Oh, that's cool. And then, so when did you make the transition from, you know, IT guy for the politicians to a job where you are actually working for a technology company?
Mike Conlow: So I had a kind [00:03:00] of long career in political technology, maybe 15 or 20 years.
I don't know. And so went pretty high up in the ranks of political technology, doing things like the Blast email systems that send out so much political email and the donation systems that campaigns use and the websites that campaigns use and so, so got to work on all of those very fun things over a lot of years and eventually a couple of years ago, kind of.
was ready for new and different challenges. And so started getting involved in the digital divide and research around the digital divide and how we can kind of get every American connected. And that kind of fortuitously led to a job at CloudFlare where I've been very lucky. find a kind of culture of other technologists, other kind of tinkerers, very curious people.
And so it's been a really great fit, you know, kind of a second career still in technology, but definitely, definitely different in a lot of really [00:04:00] good ways.
Matt Trifiro: What does your job consist of? at CloudFlare.
Mike Conlow: Yeah, I'm on our infrastructure team at CloudFlare, which is a team responsible for kind of building out our network.
All of the edge locations that I'm sure we're going to talk a lot about, all the data centers, measuring the performance, how fast is the internet, how fast are our websites and our services and our. Zero trust and sassy products. And I also spend a good amount of time kind of given my background on watching kind of the internet policy landscape, what is going on in the world of policy globally that might affect the internet and might affect cloudflare.
Matt Trifiro: Yeah, wow. Okay, we'll definitely cover both those topics at length. So, when you think of companies like Cloudflare, and I'm going to put this in air quotes, a CDN, because you're not, you know, caching is something you do, but it's been a long time since all you did was caching. So actually, why don't we start there?
Why don't you give the listeners a sense of, the range of services that someone can purchase from CloudFlare and
Mike Conlow: [00:05:00] how that works. Right. I think you, you hit on it. CloudFlare kind of started as a CDN, make websites faster and more secure. But in the last few years, bunch of years, we've really expanded into a bunch of other areas.
And so we, we would call the CDN kind of our application services. Another big group of products is our developer platform. We call that Workers, and it's an Edge compute platform, and I'm sure we'll talk a bunch about that. But, you know, it's a set of services where you can deploy a webpage out to the Edge.
You can store your HTML and your JavaScript and your images all out on the Edge. And then, with the Workers platform, you can run code at the Edge. And so that's in our kind of growing developer platform stack, which is very quickly becoming kind of a full stack that you would need to deploy any kind of application.
And then there's a whole suite of kind of newer network services, net network as a service network products that are run through our edge servers. And those are kind of, they fall under kind of the general sassy [00:06:00] bucket and kind of sassy can mean a lot of things to a lot of people, but I think of it as.
It's a no VPN solution. It's a zero trust solution. So I'm connected to internal cloudflare resources, but I'm not connected to a VPN. And so all of the security, we can run a, we can run a gateway that's monitoring traffic going outbound. We can use, we can have a firewall that's protecting your traffic.
We have a product called Magic WAN that connects lots of different clients like laptops, whole data centers, remote office buildings can all be connected through a software defined WAN. And then we put our security services on top of that so that not everybody can see everything else in the network.
You're kind of, you're limited. By what you can access on that way in and that's our magic way in product And so there's a whole suite of these sassy these sassy products that we offer all together It's not a whole bunch of different vendors a whole bunch of different solutions. You can really get The entirety of the SASE [00:07:00] portfolio through CloudFlare now.
Matt Trifiro: That's great. Now, it's kind of funny because you said the way you define SASE, right? And so, first of all, you mean S A S E not S A S S Y, which is the name of the stuff logo that Mark Benioff brings to Drainforce. SASE. This is S A S E. What is the canonical definition of SASE and how do you differ from that?
I think you need to
check
Mike Conlow: with Gartner on, on the canonical definition of, of, I think of it, I don't want to repeat my previous answer, but I think of it as a way to run security services. From a cloud from a edge network instead of a whole bunch of on premise equipment. That is Monitoring your traffic and blocking certain things and protect.
I have
Matt Trifiro: all of that on premises when it can be delivered from right? Exactly.
Mike Conlow: There's edge exactly, right? And and so we'll talk more about it. But once you have a very complete edge network globally All of a sudden running all of that [00:08:00] logic about what is attack traffic and what is not, we can deliver that in a very performant way from our edge network.
And we don't need to have a whole bunch of very expensive, very, very fancy on premise boxes doing network. We can do it. with computing power on our edge.
Matt Trifiro: Yeah. Okay. Now let's transition a little bit to what Cloudflare looks like in the real world, right? So can you help us understand the size of your network?
Like how many POPs you have? Just give me a sense of how big Cloudflare is.
Mike Conlow: Yep. We're, we have a presence in 85 cities globally. I don't want to scoop anybody, but we are coming up fast on 300 cities globally. And, and each of those cities are places where we connect with other networks and have computing power.
So that is our edge network. The way I like to think about it is we want to be really close to eyeballs, to end users, to anyone who's accessing the [00:09:00] internet. And
Matt Trifiro: maybe not even people, devices, sensors, right? Right. Yeah, okay.
Mike Conlow: Right. And so we want to put an edge location, a pop in, in as many markets in the world as we can.
And then what happens is hopefully no matter where you are. In the globe, your request for data for an API service goes to your ISPs, comes from your house to your ISPs kind of aggregation hub, and then hopefully the next step after that is it finds CloudFlare at a data center. In your market, and so the latencies that we're talking about, there are sub 50 milliseconds when we're, we're in the same market as, as you and actually quite a lot, a lot below that cloudflare is within 50 milliseconds of 95.
of users of the internet. And so now I
Matt Trifiro: imagine that's a, that's a real number that like you actually [00:10:00] test the performance of your network, because there are other companies that shower me nameless that say things like, Oh, we're three milliseconds from everybody on the planet. And the reality is they're not, cause they don't have that last mile access network that gets to the, as you call the eyeballs or the people or the businesses or the devices, right?
So 50 milliseconds is where you. I mean, and below, but it's ones in tens of milliseconds to hit cloudflare, where it might be hundreds of milliseconds if it had to home all the way back to some server that you were. Exactly
Mike Conlow: right. Exactly right. And so once we have all of these data centers all over the world, yeah, your, your request for traffic is coming from your house.
To your ISP and and then hopefully it hits Cloudflare in that same market at a neutral data center where your ISP has a router and Cloudflare has a router and and compute or maybe Your ISP hosts a embedded cache from Cloudflare the same way they do from Netflix or a bunch of other companies. And so maybe [00:11:00] your request for our services is in the network, right?
And so we do both of those things. But once, once that is all set up, everything is just fast. We have to work on the network with the rural areas. We have to work on the networks that we're not directly connected to. We need to still monitor and make sure we can increase performance where it's not as good as it as it could be.
But we can get to a lot of people very, very quickly. Thank you. If we have a broad kind of edge deployment and connect with a lot of other networks. And so CloudFlare is connected to over 12, 000 other networks on the internet, which is a lot of, a
Matt Trifiro: lot of networks. That's a big number. Yeah. And how many of those almost 300 are in the United States?
Roughly. Uh,
Mike Conlow: 50 ish, something like that. We don't have one in every state. We have one in a lot of states. Is it more
Matt Trifiro: or less the top 50 cities in the U. S. population wise, with some exceptions of like, that's where a big interconnection point is or something? Okay. That makes sense to me. That makes sense to me.
[00:12:00] Okay. So, when I think of, and I'm just going to focus on the United States because it's easier for me to picture in my brain. If I think of, cloud flares. network, right? And all imagine 50 pops all over the place. As I understand, I just learned this about Cloudflare that you actually deploy your full stack, like all of your services in every one of those pops.
That's accurate. Yeah. Okay. All right. So you've got these little Cloudflare's. So you said it gets faster, right? And there's fast and then there's really fast and then there's impossibly fast, right? So if you look at what's happening down at a CPU, it's nanoseconds, right? No network is probably going to be nanoseconds unless it's like embedded in a machine.
Uh, and then there's hundreds of microseconds, let's say, or tens of microseconds. And that is kind of land speed on the premises. Or near premises, so within 5 to 10 kilometers from the premise. Hundreds of microseconds. Now you were talking 15 milliseconds, 20 milliseconds, 15 milliseconds. Is it accurate to say [00:13:00] that you keep pushing your network out farther to the edge?
Is that essentially what's... And I guess more places in between, because it isn't just, like, it's this perimeter. Tell me how your network is expanding. Yeah, well,
Mike Conlow: let me give you, let me give you an example. This, the southern part of the United States is not as well covered by data centers as the northeast United States.
And so, a lot of traffic historically in the south has flown to Atlanta or Dallas. And a couple years ago, the city government in Montgomery, Alabama, decided they were going to make an investment in bringing more internet infrastructure to Alabama, and they started an internet exchange and an internet exchange is a place where the ISPs connect on one side and the content and services networks connect on the other side and they literally call them meet me rooms because the ISPs are connecting to the content network and everybody's exchanging traffic and it has a little bit of a chicken and egg problem.
You got to have [00:14:00] enough networks there that there's Traffic to exchange and the city government really took a lead in getting everybody to come, including CloudFlare. And so now CloudFlare has a presence in Montgomery, Alabama, and now subscribers to certain ISPs and universities and other big networks in Alabama.
Instead of their traffic going to Atlanta or Dallas, their traffic will find CloudFlare in Alabama, and it'll be a lot faster. Yeah,
Matt Trifiro: and so... So let me help explain to my listeners how this kind of works, because in today's world, we want to, want to get to CloudFlare and CloudFlare, let's say, is at the regional data center.
And because it's not this investment that this community made, and let's say the regional data center is 300 miles away from where I am, and I want to go to another network that's in my local area, but in order to do that, I have to go back to the location where. Cloudflare is and where these servers are.
And the term of art is sort of hairpin, which [00:15:00] means that you have to go out to somewhere to get what you need and then come back. And I guess the original internet was the ultimate hairpin. You had to go all the way back to the centralized data center. And you've sort of pulled the top of that hairpin down closer to the thing.
And it seems natural that the more. Interconnection points, the more meet me rooms, the more IXPs, the closer they are to the access networks and the people in the eyeballs, the better because then there's less hair pinning. And it sounds like and I certainly know this from experience to that. This municipality in Alabama isn't the only group that's trying to set up a new interconnection points.
And so I just want my listeners to understand that part of. Okay. What happens when you push your network out farther is you start creating gravity for more interconnection points, more meet me rooms, overall general performance of the Internet should go up. That's right. Yep.
Mike Conlow: Very well said.
Matt Trifiro: Okay, so let's, let's talk about Edge, right?
How do [00:16:00] you think about... I think
Mike Conlow: we just covered a bunch of it. We want the edge. We want to push out the edge. Everything we were just talking about is the edge of our network. We want to push that out absolutely as far as we can, and we want to make it so that whatever ISP you are on, that you can find Cloudflare ideally in your own market.
Matt Trifiro: Ideally in that network or adjacent to that
Mike Conlow: network. Yeah. Right. Well, and in all 12,000 of those networks that we connect to, we, we are adjacent to the other networks. And so yeah, it's even better if we can embed within the ISPs network, but I don't think we should poo PPO at all. The, the idea that we can reach an is p in a market in a neutral data center with peering, that's, that's a great way to exchange traffic when we can do it locally.
Edge gets thrown out a lot, obviously, and so I would compare. That to a system that calls itself edge, where you are deploying [00:17:00] your application into a specific region or zone or something. And I've gone through some of these with other systems and the instructions on how to deploy your application into one specific ISP hosted zone are pages and pages long.
And you're just deploying your application into one network, into one market. You got to set aside hours to do this.
Matt Trifiro: Yeah, it's, it's interesting when Amazon, for all intents and purposes, up until recently had two regions of the United States, East and West, right? And a human could figure out how to deploy to those two.
And if something went down, a human gets paged and could fix it and stuff. But when you have 300 plus, and at some point 3, plus, this could be super distributed. You can't do it with a human anymore. So I imagine that Cloudflare has, I suspect that. You said in contrast that there aren't 30 pages to have to go through to figure out.
So how do I deploy an app? to
Mike Conlow: CloudFlare's Edge. Right. And so, this is our kind of worker's platform. [00:18:00] You can run code, any code you want. Is it, maybe it's an API backend. Maybe it's your website. Maybe it's some new fancy 5G remote surgery thing that everybody keeps hearing about.
Matt Trifiro: Are these workloads in containers?
Is it serverless? What's the model?
Mike Conlow: It's all serverless. You're writing code, and then you test it locally, and you hit deploy, and that code is now going to go to... All of those data centers in all of those 285 plus cities, and now that code is running at all the cities and so whatever the best path is for data to find CloudFlare for a request for our content to to recover, it's going to take that same path.
It's going to be the same. Sub 50 milliseconds and probably better, and then it will hit cloud first edge and covers edge will execute the code and send it back to the user. And if you want to change the code, if you want to update your application, you make your updates and you [00:19:00] press deploy. And it deploys it out to all of Cloudflare's 285 plus cities globally.
And within a millisecond, a couple of milliseconds, depending on where you are on the world, but literally I've done it when the person's on another continent and they click refresh on their browser. And in not very long, they see the update that I've made to my code. It doesn't matter what ISP they're on.
It doesn't matter where they are in the world. They can do that. And so you can write. any application you want. It's just a, it's a developer platform. And
Matt Trifiro: when I deploy, right, I push the deploy button or whatever the command line is for that, rather than saying, okay, I want this service to run in this specific location and this service to run in this specific location.
And hopefully I've got a good disaster recovery plan. Like you just make copies of it and put it everywhere. So every application that anybody writes for Cloudflare is running on every Cloudflare server, more or less. Right. Wow. So that is globally distributed edge. It is like [00:20:00] done. Right. That's very cool.
Right. And I think that one of the early misconceptions, although people don't seem to have this quite as much, but I do think it's worth pointing out because. Many people come from the perspective of the device or the on premises. And then there are other people, like you and I, more or less, that come from the infrastructure side.
And in some senses, the infrastructure edge and the on premises edge are getting closer and are about to meet. But it's really important to understand the differences of services delivered from the infrastructure. And I think one of the things that people don't fully appreciate is Cloudflare, you're essentially operating as a, as a serverless cloud provider.
I mean, that's what you are. Whether you want to admit you're competing with Amazon or not, at some point you're competing with Amazon for edge workloads, right? And yeah, that's something around a server that someone, other cloud provider might want to try to capture themselves. You've got the ability to.
Deliver services like your secure, your sassy products that used to be on premises [00:21:00] can now be delivered from the network. And so it's not just it equipment and it stuff like firewalls and things like your sassy product, but it's, it's things like smart retail or computer vision. And so you think about what helps monetize edge having cloud economics, the ability for your developers.
To be able to program something easily deployed everywhere. Okay, that's cool. That accelerates that whole process. And then, to have the ability to pay you like a cloud provider. Pay as I go. And not deploy any hardware. I can be an asset light business like most software developers want to be. And let Cloudflare take on the capital expense of deploying that equipment.
Let's imagine I'm a smart retail. Maker and I go to my grocery store chain and I say, okay, the good news. I have this amazing smart retail solution. The bad news is you have to put 500 cameras on the ceiling. You have to connect them all together. Then you have to, Oh, your telecom closet's not big enough.
So you actually need a data center in your parking lot. But once you [00:22:00] have all that. In all 300 stores, right, right. And the head pops of the retailer. But if you go in and say, okay, you have to have the cameras inside, right? But we can deploy a wireless network that can connect the cameras. And you don't need that data center because we're going to do the inferencing on the cloud.
And the really low latency inferencing that has to happen fast are going to be done by CloudFlare workers. And some other workloads might be done by the centralized cloud. And that totally changes. It's the economics of the services that you deliver to retailers or manufacturers or so on. And I'm just interested in how, if you agree with that, and if you, if you see that being an accelerant.
Mike Conlow: No, I, I definitely agree with that. I think one way to think about it is let's take an oil, an oil derrick in North Dakota or South Dakota or, or something. If there is a compute need, the, some kind of data processing. need. If the use case needs kind of sub 10 millisecond latency. [00:23:00] Okay. We're going to have to really kind of work.
We're going to have to figure that out. Maybe that needs to be on premise somehow. But I think if you think about what the use cases are, if the use case only needs really good latency, not like sitting on your site latency, then is running that workload from a cloudflare edge that is in Minneapolis or Chicago or some place that's not next door, but not that far.
Will that still work? And now what we can do is focus on really good underlying layer two connectivity for all of these sites. And the same for the retail example, focus on getting fiber optic internet to every, everywhere that might need it. And that is reliable. Where the ISP has really good peering with lots of different cloud providers and now the subscribers to those ISPs, be they a commercial or [00:24:00] industrial use in the Dakotas or retailers or people working from home who can now use the services delivered from home.
The cloud from from an edge service, and we don't need to do all that computing on premise or at the retail site. We're just moving that computing into a cloud, but it is a local version.
Matt Trifiro: Well, I mean, you know, what's that the the professor that the cloudlets? You know, that's, that's, you've essentially built the Cloudlet.
Just, you didn't tell anybody until, until it's done. As someone who co founded the State of the Edge, which is now a Linux Foundation project, that was trying to define what the Edge is and the different versions of it, and is the host of the Over the Edge podcast, I'm of the belief, and I think you're a proof point, Cloudflare's a proof point, I'm of the belief that in five years, we're probably not going to be talking about Edge anymore.
It's just going to be the internet. And it's just the internet's going to be faster and better [00:25:00] and more discreet timing and better routes and more disaster recovery so that we can run more things from the internet. And it sounds like you're nodding your head. So,
Mike Conlow: yeah, I think about that when I, whenever I think about what is the killer 5g use case, and I don't know what, what it's going to be.
I think we're all kind of, we're
Matt Trifiro: computer vision, computer vision.
Mike Conlow: Okay. I buy that as, as one, as one possible thing. One thing that I. think about it is sometimes I'm, I shouldn't do this, but I'm, I'm in my car and I'm trying to order my Dunkin Donuts on my Dunkin Donuts app and I'm driving, I shouldn't be looking at, keep looking at my phone, but as soon as I open my Dunkin Donuts app, I'm watching it do its thinking, like what are my last orders?
What are my, what are the specials today that they want to show me? And I'm thinking about how much time elapsed for it to fetch my last orders from some Distant data center where they have that story and to fetch what the special of the day is and I think what if that They could get [00:26:00] my order history from a local cache deployment What if they could get that all their specials from that same local deployment?
They're gonna need to probably go to a centralized server when I press the order button or something sure but there's a lot of the internet that I think will just get a lot faster. So I, I think that you're exactly right that we should make part of the story about 5G and about edge computing, that just the whole internet is going to get a lot faster to the point where it should feel kind of mostly instantaneous.
Matt Trifiro: Yeah. And that's, that's another interesting distinction that I like to draw because I think it helps people really kind of wrap their head around this is that we live in a world today. And historically. Where the primary use of the internet was humans like you on your Dunkin Donuts app, communicating with some server that's out somewhere else, right?
And yes, I'm just as frustrated by the one, one thousand, two, one thousand, you know, the five seconds it takes to get, like, I want that instantly. I totally agree with you. Better experience. And sometimes I'll abandon the cart. Like Dunkin Donuts will make more money if they can make [00:27:00] that, that service faster.
That's like Amazon's proven that over and over again. But on the scale of machines, So on the scale of running a 5G RAN or the scale of running a real time computer vision inferencing where now you're actually operating in milliseconds and microseconds, not ones of seconds, because you're talking to machines, not And so I think we're crossing that point where most of what's going to happen out at the edge may very well be machines talking to machines.
And I think that's very, very likely. I'm interested if cloud hair flare up in opinion or if you have an opinion on that.
Mike Conlow: No, I mean, I, I agree. We're, we're not super opinionated about who is talking to what, right. And so if it's, if you have an API service that you're using to, you have a fleet of devices and you have a API service that controls them and takes data in from them.
Please run that on our platform. And if you have a website that's consumer focused, that works too. We want to be a [00:28:00] platform for any application. I, one thing you said made me think of something, and this is an example from an IETF paper. I scrolling on. On Google Maps, you're on Street View and Street View has to download a bunch of extra data about where you might scroll next, and so that's going to use a whole bunch of bandwidth.
That's going to consume data, but if we can get these networks to the point where I can scroll on Street View and We can get that request to, to Street View very quickly and Street View can respond with just the new frames that I need that I requested. Now it doesn't need to download and store a whole bunch of other frames that I didn't request.
It can really kind of interact more instantaneously. Instead of trying to predict what I may scroll to, it can just download the frames that I actually scroll to. And so I think there's efficiencies. That
Matt Trifiro: makes a lot of sense to me. It makes a lot of sense to [00:29:00] me. There's going to be some trade offs between what you run on the infrastructure edge and what you run on the device edge.
But one thing that's clear is even though we tend to think of the infrastructure edge as being constrained compared to the centralized cloud, it's a lot less constrained than the device. And so while some things might be running on the device, you're ideally going to have a lot of. Things running on these scalable environments, these cloud environments that are out at the edge as opposed to, because I'm not going to run a cloud on my device unless it's a server.
I mean,
Mike Conlow: don't we all kind of almost use our computers as clients? Web browser. Yeah, my
Matt Trifiro: computer is just a hardware web browser. Right.
Mike Conlow: I have a very, very powerful client for the internet. Yeah,
Matt Trifiro: that's really true. That's kind of funny when you think about it. Let's transition a little bit. And. Let's talk about the digital divide.
So you said that one of the things that you're passionate about. Is identifying is kind of your political background is paying attention to things that, I guess, on a [00:30:00] policy perspective, have an impact on the Internet and us as users of the Internet. What are the things you'd like to draw to our attention?
Well, I think
Mike Conlow: in terms of digital divide, we should think about it. There are still people about the Internet. 11 or 12 million of them in the, in the United States that really don't have access to adequate internet. Meaning that they don't have a wire outside of their house or, or a wireless way to connect to really good internet.
That's one issue. And there's a lot of attention on that and we should talk, talk about it. There's another piece of this that's important to keep in mind, which is there are a bunch of other millions of people about the same number when you kind of do the math who do have. They technically have access to the internet, but are not using it.
They either can't afford it, they don't know how to use it. That is a, a kind of separate problem. But when we think about digital divide, we should think about both of these things, both people who can't access the internet and people who aren't accessing the internet. And what can we [00:31:00] do to get everybody connected to the internet?
Not just wires in front of their house, but, but that's important, but also how we get them on the internet and using the internet to its full, its full potential.
Matt Trifiro: So tell me about, I mean, what you'd like to see happen.
Mike Conlow: Well, I think I, I think actually, and this gets into a little bit of the policy between the United States and, and other countries, but I think the United States is on a, a great path.
The federal government and the United States has put, probably, if you add it all up, probably a hundred billion dollars is now flowing to this problem that the access problem, in particular,
Matt Trifiro: a hundred billion dollars over what period of time.
Mike Conlow: Over the next bunch, a handful of years. And so the infrastructure bill put 42.
5 billion directly into deployment of broadband networks. That process is kind of working its way. Through right now, the FCC, just the other day, Wednesday, I think, released a new version of a map of who has broadband and who doesn't, that the [00:32:00] Department of Commerce will use to allocate the money, and so that is happening.
There were a couple of different kind of COVID relief bills that put a bunch of tens of billions of dollars into broadband. The FCC has other programs that put. Money to broadband. There are a lot of programs right now that are putting money into broadband deployment. There's an important program for low income families that get a $30 a month subsidy off their broadband bill, which will run out of money at some point if Congress doesn't extend it called the Affordable Connectivity Program.
There's a lot of attention on this right now, which is great.
Matt Trifiro: Yeah, I mean, I. I'm spoiled, obviously, because I'm not one of the 11 or 12 million who doesn't have access to the internet. I've got really fast internet, and I know how to use it. You'd be a real disadvantage in everything in life if you didn't.
I mean, I make my doctor's appointments. I, I, like, everything happens on the internet. I don't even use the phone if I don't have to, like the voice phone. The phone app doesn't get much use on my
Mike Conlow: phone. I'm the same way. I live in a [00:33:00] relatively higher income area. I have three wired broadband options that puts me in the 6% of America that has that.
And so when you have three wired broadband options to choose from two cable networks and one fiber network, there's a competition and price thing that happens. And I probably. I pay 40 a month for fiber optic internet. I probably pay less in for internet than most of the country. That doesn't make a whole lot, a lot of sense.
And so part of this conversation should be everybody needs to get access to, to the internet. And then, and yeah, the second part of what we talked about is we need to make sure that everybody is able to use that, that wire in front of their house. Why is there
Matt Trifiro: such a lack of competition in many local markets?
Cause you had 6% only have three, you know what I mean? You have two, what percent have two choices?
Mike Conlow: I don't have it in front of me, but it's like 50% that only have one and 12% that have zero. And the balance are, yeah, [00:34:00] something like that. So a
Matt Trifiro: lot of people only have one, right, right. Why?
Mike Conlow: Well, it's pretty expensive to deploy these networks.
So the first network that deploys gets 80 or 90% of the people. They're going to use their service for broadband. And a second network that puts in a lot of capital to deploy. And they, so they have to steal customers. They steal customers from the first one. And then the business case gets really hard for the third network to deploy.
And the networks that do deploy, I've spent a tremendous amount on capital. Very careful about how they deploy their capital. And so I think what we're seeing now with. The broadband money that the federal government is putting in is in places where there hasn't been the business case, or it's, it's just too expensive to lay all that fiber that the federal government will help with the capital cost of laying all that fiber to get people access where the, where the market hasn't done it [00:35:00] by itself yet.
Matt Trifiro: So one of the things that I've seen in the wireless industry is that the hand wringing about. Well, gosh, I mean, in the U. S. there's three, four, if you count major cell phone providers, right? Depending on how you count, three or four. And they all overbuild. You have competition in almost everywhere, where I think, and 5G and 3G and 4G networks are expensive to deploy also.
So somehow they figured it out, but they haven't really figured it out because when you look at getting down to like the densities of millimeter wave, where you've got to have these small cells every block, basically, that just seems pretty expensive. And so the creative business solution that seems to be emerging, but I don't know if it's going to actually work because it requires a lot of cooperation, is neutral host.
So the idea that one company could come in and fund the infrastructure and then lease it, kind of like a cloud provider does, lease, lease the capacity to others, which happens inside, like in malls and stuff today already, like if I'm Verizon [00:36:00] and the mall provider put up a neutral host network and I can connect to it now, Verizon customers have great cell phone coverage inside this giant mall.
And I wonder, I just wonder if you, if you've had any experience with, with like sharing infrastructure as opposed to like overbuilding.
Mike Conlow: Well, in the mobile world, I'm a big fan of the MVNOs, and I think we have a, we have a good MVNO ecosystem. Can you explain what that is? Yeah, MVNO is a company that doesn't own the towers, that doesn't own the cell towers.
They are kind of leasing in a wholesale capacity from... One of the big three carriers in the U. S.
Matt Trifiro: Virgin Mobile, Mint Mobile, those are all MVNOs, right? They're just running on someone else's network and paying for it.
Mike Conlow: Right, and they have a wholesale agreement with one of the carriers and they sell a retail service to customers and it adds choice.
And I think we're lucky to have that. Is that driven
Matt Trifiro: by just free market or is that? My
Mike Conlow: understanding is that it is a [00:37:00] mostly a free market thing in the United States. Because that's kind of
Matt Trifiro: crazy progressive for someone in a large carrier to go. I want to enable my competitors from a consumer brand standpoint.
Right.
Mike Conlow: I think I agree with that. And I think part of it is that we do to your point. There are three capable. Wireless MNOs, real actual networks with actual infrastructure, they have to compete. If one of them offers this service, then they could be an M V N O host. It would take away from the other one.
So if we only had one wireless network, then I'm not sure we would have this situation, but it's a, yeah, it's a great place to be. Um, I'm sure they've thought of that, that their MVNOs could take their customers, they could also take. Customers from their two competitors, but whatever, whatever it is has happened.
Oh, that's true. Right. The fixed version of that. is an open access network. And in Europe, they are more common. Canada is working through it. There's some kind of growing pains with it about [00:38:00] what prices should be charged. We have some examples of this in the United States. Utopia, based in Utah, is a kind of growing open access network.
There's a couple of other examples. It's not super widespread. I would love for it to grow. Sonic is a network that runs on top of other infrastructure in a lot of cases. And so this kind of thing does exist in the fixed broadband world, but it is not super common in the United States.
Matt Trifiro: Yeah. Wow. That's, that's really interesting.
So when you look at the overall advancement of the internet, Right. And let's not just make it specific digitalized. So part of it is we need to get the internet in the hands of more people and make sure that they know how to use it, that there's training and education and these like large programs that put that together.
What other things should we be doing to improve the internet?
Mike Conlow: I think we're actually in the United States. I think we're pretty lucky. I think we have a really strong internet ecosystem in the United States. And so part of [00:39:00] my. answer on this is, is what not to do. I think that there are some things, there are some things not to do.
And so I'll give you one example is South Korea has essentially regulated wholesale bandwidth pricing. And so what, what happens in South Korea is that the idea that we've been talking about where a request for content and then serving content and services from the closest. place totally breaks down because the price to send data to the major ISPs in Korea, it's like 30 times the price to send data across the internet in North America and in Europe because the ISPs think that they should get these much higher prices if, if they have received your data that you should.
Pay them for the privilege.
Matt Trifiro: The operators in the U. S. have tried to do that too. And some types have been successful. I mean, Comcast with Netflix years ago. I don't know if you remember that. Yeah, right. I
Mike Conlow: do. I do. I [00:40:00] think that the market has to kind of figure these things out. I'm not necessarily saying that all traffic should be free between everybody.
But I don't think that government should get in the business of saying what the price is for one network to send data to another network, and that's, that's the case in Korea. It means that in some cases, somebody in South Korea who is requesting content and services, their request for that data could, GoFromKorea could pass the content in our data center in Seoul and then travel across the ocean to Los Angeles Where it gets served from and then all the way back to Korea, which is not a good because it's cheaper right because the wholesale prices in the United States are competitive whereas the price that the ISP wants inside of Korea is 30 times that.
And so I use that as an example, but it is unfortunately, at least in my view, now in Europe, they have taken this and said, [00:41:00] well, we like that. We like what they have going on in South Korea where we want to charge content and services to send data to our. ISPs, and they're just targeting the kind of biggest tech companies, but the idea is the same.
If you want the pleasure of fulfilling your subscribers own request for data, your subscriber requested the content from Netflix for Netflix to send the stream back that the subscriber requested that. That they're supposed to pay to the telco. It's a gnarly
Matt Trifiro: problem because there are real costs involved and everybody's trying to make margin because they've all got these capital expenses they're trying to pay off and interest rates are going up.
I mean, to some extent, because there's so much. volunteer cooperation on the internet, or at least like contractually negotiated cooperation, which is volunteer. That's amazing it even works at all. It's a little bit like using the paid version of ChatGPT. It's like, this is going to cost you [00:42:00] six cents.
It's like, do I want to get that web page or not? That's a weird world to live in, but there's a cost there. There's a real cost. It's super interesting.
Mike Conlow: I think in the United States, we're working on a solution to this. The federal government is putting up money. They're saying that there are, there are places that are too expensive for ISPs to reasonably build out infrastructure.
And we, the federal government, we're going to help. Whereas it seems to me that in Europe, the idea is that by charging. Big American tech companies, a little bit of money to deliver content that the ISPs would then put that money into the infrastructure. I'm not, I'm not sure, but if that, I think there's, there's going to need to be more of a plan than, than that to get to reach the hardest to reach places with fiber.
And I think the United States is absolutely on the right track with the federal government helping out.
Matt Trifiro: Yeah. You know, you know, one of the things that's, that's interesting to me, and I was [00:43:00] talking to someone who shall remain nameless, but this is a person who. Has built big parts of the internet, so it would know and he said, most people don't realize this, but probably 75% of the traffic on the internet.
It's not actually on the public internet. It's on private networks, which is partly due to the fact that a lot of the companies that have to deliver content have built their own networks. Google's built their own network. Microsoft built their own network. Comcast has a national network. Does Cloudflare have its own network?
Yeah. So
Mike Conlow: all of these, not all of, but many of these edge sites have, we have our own backbone that connects them to get together and it connects them to our core data centers where, where we process logs and all that kind of, all that kind of stuff that's that's not very edgy, but we have a backbone network that allows us to send data to ourselves, right?
And when we connect to other networks from these edge sites that that's an external connection, but then we have a [00:44:00] wire, either a dark fiber or a kind of lit wave that connects one site to the other and allows us to send data between, between our edge sites, from our edge sites to kind of a processing core.
And so.
Matt Trifiro: Do you use your national network, not just for like the command and control and updating of your own? Pops, but for customer traffic. Sure. Yeah.
Mike Conlow: I mean, customer traffic flows across that. Like, let me give you an example. Magic transit is what we call it's our kind of layer three denial service protection, so it allows you to, you have an IP address.
And in the kind of old internet, that IP address is on the internet. If somebody wants to attack that IP address, they can very easily overwhelm it because it's, it kind of exists at one place on the internet. But with Magic Transit, we will take your IP address and we will kind of advertise it from all of our edge servers globally.
And so now if somebody tries to attack that IP address, it's being advertised from all of our edges. So
Matt Trifiro: you just shut down the nodes that are getting beat up and
Mike Conlow: No, it's easier than that. We'll take, [00:45:00] we take in the traffic kind of gets dispersed out because if you have a
Matt Trifiro: global balancing,
Mike Conlow: yeah, you have a kind of a, somebody has taken over devices that they're using for these DDoS attacks and they attack this IP address.
But now that IP address is all over the world. And so Our edges see a little bit of that. Each of our edges see a little bit of that attack traffic. And then what the edges do is they evaluate it and see whether it's an attack or not. And the good traffic we are going to transit. And that's why it's called magic transit.
We are going to transit it. From the edge that caught it, the edge that saw it originally. And again, this is the good traffic. We're going to transit that traffic to its real origin, where the actual or origin website is, and then send it back. But that. From our edge that saw the traffic originally to the real origin, we could reach with a transit provider that can reach the whole internet, or we could use [00:46:00] our own backbone to do that.
And we do that a lot is we will take the traffic in a kind of 1 of the outer reaches, but then the origin server is in a major market. And so we need to. go from the edge to the major market, and we can use a backbone link.
Matt Trifiro: So is it true that let's call denial of service security because I don't really know what else to call it.
But for those types of service, is it true that the more POPs you deploy, the more robust it becomes because of that fact that like it distributes it across all the nodes? Right?
Mike Conlow: Yeah, absolutely. Because the more POPs that are deployed, the more distributed the attack becomes. I suppose in theory, you could Take, if in the alternative scenario, Cloudflare had one pop, we would just need a tremendous amount of capacity and processing power and everything to absorb the entirety of that, of that attack at that one pop, it seems better to distribute out that, that attack across the whole world.
Matt Trifiro: Yeah. Wow. [00:47:00] That's neat. All right. So let's look to the future. What are you most excited about? Well,
Mike Conlow: I bet you get this answer a lot, but it's hard not to be at least a little bit excited by what's going on in AI as someone who writes a little bit of code, but not that much. I can put, if I want to do something in code now, I can put it into chat GPT and it'll spit back out something that's pretty close.
Maybe it's not exactly right, but it's, it's, it's pretty close. And so for the minute, I'm not worried about chat GPT taking my job. I think that chat GPT is making me a lot more. A lot faster at, at my job because I don't spend so much time figuring out syntax and how to move the legend on the chart into the upper right corner,
Matt Trifiro: you know, look up how to use VLOOKUP anymore.
Like what order the parameters are in. I just click hashtag to do it for me.
Mike Conlow: It's hard not to be a little bit. Interested and excited about where that's going. I think I'm, I'm pretty interested in, in where the kind of IOT space is going. Some of the stuff we talked about, if you can deploy [00:48:00] sensors that, that for agriculture and a lot of these things that are kind of in use now, or starting to be in use now, but the price needs to come down and the adoption needs to go up, but we can deploy sensors for agriculture uses, but also for how often the water your garden, how often the water your lawn or.
What parking spots are free? Cause there's a sensor in the parking spots. Oh my God. Those
Matt Trifiro: parking garages that have those signs and green and red lights are
Mike Conlow: amazing. Right. And so I think, I don't know what all of the uses of this are, but I think that there's going to be more and more of these kinds of internet connected things that are just going to make things faster and more efficient and has sustainability impacts and makes our lives easier.
And I'm excited about.
Matt Trifiro: There's a cliche or a trope, but I still like the sentiment of it. And that is all great technology becomes transparent, right? And that's what you want. You like, you don't want to think, Oh, this screen on my Dunkin Donuts is taking longer to load because. [00:49:00] You just want it to be as instantaneous as that.
Yeah, that's it. That's a neat thing to look forward to in the future. What about from a policy standpoint? What would you like to see in the future from a policy standpoint?
Mike Conlow: Well, again, I'm, I'm relatively happy with where things stand on internet policy in the U. S. I think that we have a healthy internet ecosystem.
Obviously, I'm a supporter of net neutrality, but I'm not sure that I think that we're in a very bad place currently with how things work. I'm not kind of rabid about what we need. Net neutrality passed right now. It seems to be that like it would be a. Big fight and I'm not sure that all that much would change.
So I'm, I'm pretty happy with where things stand right now. I, I obviously want to see this broadband money put to good use. I want to see that remaining 12% of people who don't have access to good internet. Get it. And I, I, I think that this is, it's, it's a bit of a cliche too, but this is kind of a once [00:50:00] in a generation moment in terms of funding that's available.
And so I think we will be in trouble if we don't get it done with this batch of funding. But overall, I'm happy with where, where things stand on, on internet policy in the U S I think we're, we're, we're doing well.
Matt Trifiro: Yeah, that's great. Well, Mike, this has been a delightful conversation. Really appreciate having you on the show and sharing everything about.
Everything, all your knowledge about digital divide and how CloudFlare works and where the edge is. So thank you very much. Yeah. Thank you, Matt.
Mike Conlow: That does it for this episode of over the edge. If you're enjoying the show, please leave a rating and a review and tell a friend over the edge is made possible through the generous sponsorship of our partners at Dell technologies.
Simplify your edge so you can generate more value. Learn more by visiting dell. com.