Jan 12, 2023

Simplifying Global Networks + Insights Into Startup Investing

In this episode, Jake Warner chats with Tom Daly, CEO at Big Network, and former Senior Vice President of Infrastructure at Fastly. Discussions include his early days as co-founder of Dyn, experience with the build-out of Fastly, and his philosophy behind startup investments.

Transcript

Jake Warner + Tom Daly



Jake Warner: [00:00:04] Hello everyone, and welcome to the 16th episode of Cycles podcast. I'm Jake Warner, Cycle CEO and founder. And with us today is Tom Daly. Tom Daly is the CEO of Big Network and before Big Network, Tom was the co-founder at DYN, which was acquired by Oracle and has also previously held a senior leadership position at Fastly. Tom and I have known each other for a couple of years now, and as of last year, Tom became an investor in cycle. Tom, it's great to have you on our podcast today.

Tom Daly: [00:00:32] Yeah. Thank you, Jake. I appreciate the opportunity to meet your audience and speak to them.

Jake Warner: [00:00:36] Excellent. Excellent. So I've been looking forward to this podcast for a long time. You and I met let's see, it was about a year and year and a half ago, almost two years ago now, as I was going through a raising a round and you and I were introduced as you were helping to do, helping to do due diligence, that's a tongue twister on Cycle. And, you know, since then, obviously, you've become an investor in our company. But the other thing that I've really appreciated of having that connection is the fact that you are also very, very, very, very technical. And so I guess first, as we introduce you to our audience, I like to begin with just sharing some a little bit more background on you. So you've been a founder, you've been an investor. What's everything to Tom Daly?

Tom Daly: [00:01:33] Yeah. So if we go way back in time, I started in the industry as a network engineer in well, actually, I take that back. I started in the industry in dial up technical support. For a dial up ISP. And I don't know to what extent your audience knows what dial up is, but go back to the late nineties, early 2000s, and it was the era of Windows 3.1, trumpet windsock, and on a really advanced user machine we would be configuring dial up networking on Windows 95. So, you know, to this day, if you need someone to configure dial up networking on Windows 95 for you, like I'm your guy, I'll walk you through that, right? It's just double click my computer, double click, dial up networking, double click on Add New Connection and we can walk through the rest of that from there. This still haunts me. If you need access telephone numbers around New Hampshire, I probably still have some of those memorized for local towns as well. But yeah, so I cut my teeth in the era of dial up and worked for an amazing mentor who somehow trusted a kid in high school to do a lot for him. This gentleman's name is Gent Cave. He co-founded an ISP called G4 Communications. Eventually that was merged into or acquired by First Light Fiber, who's a large operator here in in the New England area.

Tom Daly: [00:02:54] And they spun out into upstate New York as well with their network. So I had a tremendous opportunity to learn a lot. I got to learn, obviously got to learn, dial up networking, got to learn customer service. Gen trusted me to migrate the company's billing system from a bunch of flat files and CGI scripts to a SQL backed database system. I have no idea why you trusted a kid in high school to do this for him, but we did it and we brought a lot of automation along the way. So we were now, you know, provisioning users in RADIUS servers in an automated fashion. We could live test credit cards using authorized net over the Internet versus the old way we did it, where on the first day of every month you went over to one machine and copied a bunch of flat files to a floppy disk and you went over to the credit card terminal, put that floppy disk in and read all the credit card numbers off and charged customers and then got the result file and put it back on a floppy disk and brought it back to the to the ISP server to say that these cards had been successfully charged and these people can keep accessing the dial up network.

Jake Warner: [00:04:04] So you were running a legitimate “Sneaker Net”

Tom Daly: [00:04:07] I'm for the credit card data. Yeah. I mean, you don't want that on the ISP servers for sure. Yeah, it was. It was air-gaped, right. So, you know, then got really lucky and got to see a bunch of evolutions of the internet. Right. So dial up to ISDN for a whopping 128 K connection to DSL to T1s, T3s, frame relay all the way up to ethernet services and we learned a lot about BGP routing through networks. Started to understand how the DNS system worked and generally just had a blast building this ISP. Got a lot of experience in the physical layer as well. Learn that. Learn that you don't want to put a pallet of 12 volt car batteries on an elevator all at once, like it was like a 2,000 pound pallet of batteries. Don't. You may think you're on an industrial elevator, but if you haven't checked it for weight and you exceed the weight of the elevator, when you when you drag the pallet jack into the elevator and the whole elevator goes down below sub floor and it won't go up, and now you're going to load these 150 pound batteries one by one off the elevator like you have a good learning experience in that day.

Jake Warner: [00:05:31] Hopefully you're on floor one.

Tom Daly: [00:05:34] We're in the basement. Fortunately. Yes. Yes. But then. Then we got to learn, you know, like carrying like four of these up on a hand truck at a time so that we had no elevator problem. So have dealt with the physical layer as well and just, just have a blast building networks. Along the way, connected up with the team at DYB DNS back in the day. Tim Wild, Chris Reinhart, Jeremy Hitchcock and, you know, had the excitement of building a software company in the service provider space. Dyn was deployed to well over 300,000 subscribers across the world at base levels. We built the dynamic DNS functionality. We built custom DNS so that way you could bring your own domain name instead of using one of our domain names, we bootstrapped that company and spent a bunch of time building out on products to the DNS. So we had email delivery, and email relay services, store and forward. We could be your backup MX provider if you wanted. We built one of the very first network external synthetic network monitoring products before the likes of, you know, Thousand Eyes and Catchpoint ever existed in our industry. And we even got into hosting virtual machines for a period of time on, on a platform we called the spring server and it was hard and, and it wasn't our DNS roots, and we weren't really sure if this cloud thing was going to take off because we didn't call it cloud. But we had we were one of the very first VM providers out there. We're like, This really isn't a thing we're going to pass. Let's go back to our DNS roots. So we missed offering cloud.

Jake Warner: [00:07:25] Were you using like a raw like, like Libvert set up or was it?

Tom Daly: [00:07:29] Yeah, this was like with KVM or something. No, it was. Libvert Yeah, it was wild.

Jake Warner: [00:07:35] Excellent.

Tom Daly: [00:07:36] So we ultimately decided to double down on our roots. We had had some customers. Really interesting evolution. Breitling watches the watch company, high end watch manufacturer, and I think the other was Anadarko Petroleum. Both approached us and said, Hey, we have a problem, we love your product, we love your service offering, but our CFOs don't trust you to provide our enterprise DNS for $24.95 a year because that's what we were charging at the time for all of your authoritative DNS with a web portal and API. So what they said is like, Hey, do you think you could take like that $24.95 a year and turn it into like $24.95 a month for us and just put it maybe on a different website? Call it Enterprise because we really don't want to have to go to the competition like we don't like them very much. We don't think their features are as good. They charge way too much. So just like take $24.95 a year and make it $24.95 a month. And being good engineers, we said, yeah, no way, we're not doing that. We re platformed all of Dyn DNS into the Dyn Act managed DNS platform. We built out 15 anycast sites across the world, rolled out a whole new API, whole new data replication engine and showed up to those customers and said, Well, do you want this instead? It took us like a year and a half, and they were like, No, we just wanted you to change your pricing. And we're like, No, no, no. We re architected the entire platform for you. It's amazing. Here you go. They also, they.

Jake Warner: [00:09:11] End up becoming customers.

Tom Daly: [00:09:12] Came on as they became customers. They became early customers of the dynamic managed DNS platform. But it grew. I mean, it was a phenomenal platform. We had the likes of Salesforce, Netflix and Twitter on the dynac platform. So it served a significant it has served a significant amount of Internet traffic. It was eventually acquired by Oracle in 2016. And to this day, I mean, it's almost 2023 and Oracle is still running elements of the dynac platform for their DNS services. They just they just migrated the authoritative name servers away from AS33517. Long live AS33517. We're going to miss you. And on to their own their own network backbone. So you know Dyn was super fun small company 100, 150 people Amazing environment down in the heart of Manchester, which is just north of where I sit today. And. But I wanted to do something a little bit bigger and eventually joined the team at Fastly and Fastly got into the world of content delivery. Founded by a gentleman by the name of Archer Bergmann, he was the chief technology officer at Wikia. So everyone knows Wikipedia. It's the public wiki site, the public encyclopedia, but there's actually a for profit side to that business, and that's called Wikia and it's a wiki hosting website.

Tom Daly: [00:10:39] So anyone that wants a wiki for any reason can go get a hosted wiki by by at the time Archer and his team. This is since evolved into the site called Fandom. And Arthur challenge, huge challenge in front of him. So if you think about what a wiki is, you know, it's highly cacheable content. Like it's amazing to get on to a CDN, a content delivery network because the content doesn't change very often. So you can put that on somebody's caching network all over the world. The problem is that when it does need to change, you need to push that that new content into the cache as quickly as possible. And there's really two ways, even in 2022 to do that, to signal that to a CDN as the origin. So the first option is you set a time to live on your content, which is just an HTTP header that you respond back with say, Hey, please cache this for 60 seconds, please cache this for an hour, please cache this for a day or many and support a purge verb in HTTP and you can send that purge verb out as the origin and it will refresh the cache. The challenge with this is the lower you drop the TTL, the more load you have on your origin, right? The higher you raise this TTL, the less often the CDN is going to come check in for fresh content.

Tom Daly: [00:12:07] And at the time, this was all all the way back in 2012 issuing the the purge verb on the CDNs that were out there was like the fastest network was like a 15 minute global purge, and the average was like a day. And Archer just said this isn't good enough and designed a new CDN with a sub-second purging system which fixed his problems at Fastly and opened up the opportunity for someone like me to join to help build a very, very large global infrastructure. So over six years we built 75 global points of presence and over 100 terabits of connected edge capacity to that network. So that was a blast. A few Super Bowls, Prince passing on, and and having a very major role in a variety of Web assets for the finale of Game of Thrones. Probably my most favorite evening and watching the Internet do its thing. So, you know, it was it was an absolute blast. And in 2020, hung up my hat there, decided to do a couple years tour of Angel investments. And that's where I got to meet you.

Jake Warner: [00:13:35] And I'm glad we did meet. And it is actually really interesting as you're telling these stories is like I remember, oh, I don't know, the year, but it was it was when I finally started getting interested in networking and stuff. And so this is I know I'm going to I'm going to guess 2005, maybe 2006. I don't quote me on that. But as I started, you know, it was it was the first time I had actually, you know, bought a Linksys router and started playing with it and everything. And I remember I had set up some sort of. Oh, I don't know. It must have been like an FTP server at home or something, right? Like something that I could access. I was back in, let's say, 2006. I would have been in high school. And so maybe it was maybe it was a little bit later than that. Either way, year doesn't matter. But I remember setting up an FTP server at home that I wanted to be able to access from anywhere. And back at that time I was like, Oh, I have a public IP at home. I should be able to use this.

Jake Warner: [00:14:38] And then, you know, heading to school and, you know, trying to access that public IP. And it's like, why isn't this working? I know when I left my house this morning, that was my IP. And so I so it's so interesting to hear you talk about Dyn from the side of building it where I just me being in high school, realizing the need for dynamic DNS back then I remember going into my Linksys router and choosing the Dyn option and setting up, you know, a dynamic hostname that I really just, I guess a domain that would always just constantly obviously be updated to that IP and finally being able to to set up port forwarding and getting into my FTP server. But it's just so interesting hearing you know the both sides of that is like that was like me first starting to explore that side of networking and the problems that nowadays, you know, are things that, you know, so many people are familiar with and encounter. But the early problems that we had to solve for like what happens when you have Nat in front of everything and. Rotating IPs. And all these things. So it's really interesting to hear about that. And then on that same note, as you're as you're talking about Fastly and this way of, of pushing data to the CDNs, I'm almost, this is this is me going to be nerding out a little bit, but I'm almost surprised that maybe someone's doing this nowadays. I don't know but I'm surprised. It would surprise me if no one took like as you're talking about purging and pushing up new files, if there was no like, RSYNC type method. Right. Were you just constantly looking for file changes, like you're running like an agent or something on, on some origin server. Right. And just constantly pushing out those updates versus doing this purge thing Because I mean, I think I guess me speaking or thinking out loud here is like back when bandwidth was a lot harder to come by, made sense to have these, these kind of cache events for an hour later or things like that. But nowadays, when you can just have these constant syncs between everything, I wonder if anyone actually is doing it an RSYNC, Is that something that you guys ever thought of?

Tom Daly: [00:16:47] So I know that there's some content providers and content delivery networks out there that actually do that. You know, there's kind of really sort of two models to CDN architecture in the pure content world. So there's push CDNs, right? And then there's more like your traditional pull through. So you know, a Fastly, an Akamai are more of the traditional pull through CDN. So a request comes in, it hits a cache. If it's in the cache, great. It's served as a hit. If it's not in the cache, it might be directed to a parent tier of the network somewhere or it's going to be routed back to Origin to get fulfilled. Right? So that's your traditional push-pull model CDNs. There are some there are some push model CDNs out there like Cache Fly, for example, which are really great for download traffic, for example, and they have an rsync model. So what they actually do is you sync your objects to a storage ingestion point on their network and then they do replication and they actually store every file, every object across their global network, works for smaller object footprints. But if you think of something like a streaming music service, they can easily have libraries in the tens of petabytes, the hundreds of petabytes for all of their music library. Chances are, you know, maybe 1% of that working set or 1% of that library is hot and that's your working set. So why replicate 100 petabytes of data around the world? Right? You really need maybe that one petabyte that's hot and you want that distributed around the world.

Jake Warner: [00:18:35] Yeah, and.

Tom Daly: [00:18:37] It's definitely been thought of. It's just it's just every every use case has its nuances.

Jake Warner: [00:18:42] Yeah. No, absolutely. But the reason I started thinking about it is we have some we have some companies on cycle that is really good at managing the orchestration of the underlying containers and things like that. But a lot of people are that are adopting cycle are, are working with a model where I'm sure you're familiar with like headless CMS and things, right where, where you want the data to be fully decoupled from the application whatsoever. And so for a long time people were doing this with like NFS and things like that. But in but as object storage has become more and more relevant, we have a number of companies, this is actually incredibly relevant because we had a customer working on this yesterday where they're taking an old this is like a legacy application that they're migrating. But instead of just writing to a local disk or writing to an NFS Mount, they're just treating everything as object storage in the cloud and just, you know, when a file upload comes in, it's almost just a pass through, through the front end to I mean, not necessarily a CDN but an object store. But the whole idea being there is that, that I wonder how many people are or are RSYNC-ing those types of things. And just having that native integration, which I mean, I guess to the point that you already made, if you're if you're having that like source of truth be somewhere within the CDN and they're just replicating it on their own, solves a lot of those problems while still kind of giving a single endpoint for users to, to accomplish that. So kind of really interesting how all those different things happen. So kind of a nerdy question we had we had we had Backblaze on the podcast a few months ago, and I asked a question along the lines that was how much how much data are you ingressing, you know, on a on an hourly basis or something. And I know you haven't been in Fastly for what, two or three years now.

Tom Daly: [00:20:31] Yeah, about two years now, Right?

Jake Warner: [00:20:33] Right. Two years. So back when So as you talk about the Super Bowl and things like what you know, everyone has the different definition of scale. What was Fastly's type scale, if you can remember. I know you know, it's probably one of those things that is like kind of a a toss away point. But I curious what how much is you talk about the Super Bowl and these things. How much like scale of data were you actually working with?

Tom Daly: [00:20:57] Yeah. So, I mean, I, I know the number. I won't share the number because that's proprietary to Fastly, but I can give you some sense of things because one thing that is very neat about Fastly is they're an extremely transparent company. So as you were asking the question, what I did is I went to Fastly.com/network-map and first of all, they publish their overall network capacity as of September 30th, 2022. That's 233 terabits per second of connected capacity. So that's almost like a quarter petabit per second, Right. And then if you scroll down on that web page, they actually publish in real time like the aggregate throughput on their network. So, you know, you can see this counter, it's bouncing around between like 43 and a half terabits per second and performing just a shade under 24 million requests per second. And, you know, I won't go into the specifics, but what we learn over time is that, you know, the peak over time becomes the average as you grow your customer base. So these numbers are familiar to me is what I'll say.

Jake Warner: [00:22:12] And that is and that's wild. And so I didn't expect this network map to exist. That's why I had like kind of choose a point in time in the past. But that's actually that's wild to be able to have that type of information. Of 44 and you know, 43.7 TERABITS per second right now is wild, wild. But, you know, as we talk about, you know, all those network things, you know, obviously, you know, your entire career has been focused on networks, whether it's a software perspective or hardware perspective. But I guess let's dive into what you're working on now. So you're working on Big Network. What is Big Network.

Tom Daly: [00:22:46] So Big Network is exciting. I recently joined the team over there, so. Big Network has a vision to connect clouds to anything and anywhere. So we used to have these network closets all over the place and we had these corporate data centers and we had these little data rooms, right? And eventually things moved into co-location sites. Right. That's what we put Fastly into. Fastly was in Equinix and Digital Realty and the Telex's of the world all over the place. And that was pretty standard. A lot of companies understood how that worked. And then came along cloud, and we have spent the last decade moving most of our workloads and assets into various clouds. And depending upon who you ask, depending upon the industry analysts that you ask, people believe that something like only 30% of all workloads have made it made their way to the cloud so far. And that I've heard other quotes that like something like upwards of 60% of all workloads have made their way to a cloud provider. So nonetheless, like, there's a vast amount of space in this industry for more things to migrate out of the corporate data center, out of the traditional closet and into the cloud. At the same time, we have seen this movement from the cloud providers, the central cloud providers, if you will, out to the edge. Right? We have latency sensitive applications, high speed applications, applications that are increasingly running on private networks such as like private 5G, LTE deployments that are out there in the world now.

Tom Daly: [00:24:34] And what was in these nice sort of central clouds and had local convenient networking is now moving out to the edge. At the same time, there's still a bunch of devices out in the premise that need to get connected to the cloud. And I'll share one little story about Big Network, about connecting anything to the cloud, anything was and it's sort of like got me really intrigued by the company and sort of broke 20 years of my experience. And it was actually Elon Musk's fault and StarLink. So I'm a geek. I love to try new things out. And so I got a StarLink dish and I was so excited when I got the StarLink dish. So I set it up, you know, and I always have a bunch of Linux servers sitting in the basement wherever I am because I like to be able to telnet and traceroute and ping from all over the world. And I like to be able to do it for my own infrastructure. Did I say telnet? I meant SSH. And, you know, so I get the StarLink dish and I it's got a nap box on it. And, you know, I start running some trace routes and checking what my Wan IP is and whoa, like I'm landing on a global Wan IP inside some Google infrastructure.

Tom Daly: [00:25:51] Like this is weird. And start digging around and looking at whose IP address spaces and it's like, wow, StarLink just actually does not have a lot of IPv4 address space based upon a number of their subscribers. Start digging and digging and digging and oh my gosh, my Wan IP is out of this 100.64/12 block which if you look that up in the RFCs, this is IPV4 space reserved for carrier grade NAS subscribers behind carrier grade NAT. And so I'm like. Like my entire model is broken because I'm behind this carrier grade NAT with my home Internet connection. So no dynamic DNS is going to work because it's not my IP. Even if it did port forwarding, I'm not in control of port forwarding at the carrier grade NAT box. So forget about even moving on to the next step. I sort of had to change the entire model. So Googling around late at night and checking out what the latest technologies are. And it's like go back to the normal toolbox of like, all right, well, where out on the internet do I want to establish some IPsec tunnels or where am I going to host an open VPN server? Like, I don't want to do that. And what ports? I can't open ports. So that's not even an option for me.

Tom Daly: [00:27:06] And I end up stumbling on Big Network and I tried it and it is phenomenal. It's an amazing networking technology. So think of if Napster was a music sharing app built on a peer to peer network, then Big is an Ethernet switch that's built on top of a peer to peer network. And so big provides these peer to peer discovery services from our hardware clients and our software clients. So that way, no matter kind of where you're sitting in the network, deep behind NAT carrier grade NAT, you can punch through and make Ethernet overlay or two over peer to peer at work. So. We think that there's a there's a missed opportunity in cloud networking and we're going to go seize that opportunity. And it's the space in between the cloud providers themselves where there's a lot of good players there. But between the cloud providers themselves and reaching back out into the traditional on premise infrastructure, and that can be any number of Iot sensors, building automation systems, access control systems, security systems, connecting users behind their laptops in the office to the cloud directly, and doing it without legacy technologies like IP seq. And I think for any of the network engineers listening in on the podcast, like nobody wants to configure IP seq, it is so annoying and, and we are just trying to make it really, really easy for people.

Jake Warner: [00:28:35] So with as you talk about configuring IP sec and everything. Do you remember ever having to configure I believe it was called Strong Swan or Swan Strong It was. It was. It was IP. It was key rotation on top of IP sec. I remember having to configure that one time to your point. Huge, huge, huge headache. But so with, with Big Network, there's a software component to it and a hardware component to it.

Tom Daly: [00:29:04] Right. So a couple of different pieces. The software component. The biggest software component is our control plane. We offer our service orchestrated from the cloud. So we provide a portal in a public API to our client base, and you can configure all these properties of layer 2 networks via portal or via API. So you can decide who's going to be on the network, both from devices and users. You can control different things like DHCP pools, you can configure static routing, you can do some recursive DNS manipulation, variety of options that you can control. Then once you've sort of instantiated what we call a cloud network and just think of that as like an Internet scale layer 2 domain. Once you've instantiated that cloud network, now you can start attaching things to it and there's a number of things that you can attach. We have client softwares for Windows, Mac, Linux, iOS and Android, and so those will join Cloud networks at Layer 3. And so you can build private networks, say, from your laptop to your iPhone, wherever they are in the world, You could do that for your distributed team. So if you wanted to bring back, maybe you were running a team of software engineers. When we're all sitting in an office, we were on a local LAN. That's pretty cool. Like the way I would develop with my partners at Dyn would be like, Hey, I'm running an HTTP server at 10.10.10.10.

Tom Daly: [00:30:27] Like, please come demo what I've been working on, like in the work from home world that doesn't work anymore. And there's a whole bunch of like, funny things that we're doing to our networks to make that happen. I mean, there's a bunch of people innovating in the space with SSH tunnels and or I should say reverse SSH tunnels, you know, various takes on other tunneling technologies just trying to make this easy. And you can do that with big today. One of the places that we think sets us apart is we are brave enough to delve into the world of hardware. And most folks for the past five, ten years of the world have stayed away from hardware because hardware is hard. So we actually make a couple of different devices that bridge normal Ethernet networks onto cloud networks. One of them we call Edge Lite. So I have a raw one here. It doesn't have our sticker on it yet, but it's a beautiful arm based processor with a built in heatsink. So this is one solid piece of aluminum, two Ethernet ports, one LAN, one WAN, and you can use this in remote access applications all over the place. It's powered off of USB-C, so very low power footprint. Bring in any number of power supplies that you want fully cloud managed and bridges traditional Ethernet networks to cloud networks.

Tom Daly: [00:31:51] This can be used for remote access in the data center. So if you wanted to set up an OUT-OF-BAND network, you would put the devices that you want to have remote control over off the land segment, or if you wanted to use this as an access gateway to the cloud, you could put our software up in a cloud provider and maybe you might be plugging something like a voice over IP telephone into the land port and securely tunneling up to the cloud to your cloud PBX. You might put video surveillance cameras, you might put an access control system behind here. There's just a number of options. And then for folks who want something a little bit bigger, we also have our larger multiport model. So this is six gig ports and two combo 110 gig ports available on the platform. We are also very, very close to reaching a beta period with some VM packaged gateways. So if you'd like to run our cloud gateways cloud network gateways on top of KVM or something like VMware ESXi that is up and coming functionality that we're going to be releasing. But it's just this this point of from the cloud to the edge. Let's make connectivity simple at every layer.

Jake Warner: [00:33:08] And obviously as the founder of Cycle, I get really excited about these types of things because, I mean, everything that you're doing is just so applicable to a lot of the people that are using cycle these days. And especially so, you know, as you know, and I'm guessing most people listening to the podcast, we launched our Infrastructure Abstraction Layer, or IAL, about a month and a half ago now. And so the IAL allows people to connect cycle to I mean, we have we've already, we've always had our native integrations with AWS, vulture, Equinix Metal, etc. but with the IAL now companies can come to cycle and connect infrastructure wherever it is. In fact, we actually have a couple of proof of concepts running right now with some companies. Unfortunately, we can't dive into the names of them on, on, on, on the podcast, but we've some proof of concepts running where companies now are running Cycle in their own data centers. And so it's just being the excitement of being able to reach a point someday where we can throw Big Networks devices in front or I guess in front or behind of those servers, and then also link those up to maybe an AWS account or something and have that as a seamless network as you're describing. There's just so much power there and there's so many issues that that can help mitigate because as you're talking about out of band servers, right like Cycle, as we as the CycleOS starts up, it always tries to find an out-of-band network that it can use to build all these connections, right? And if someone's running something on premises and something and they're also running something on AWS at the same time, like, you know, like unless you're setting up a cross connect and all these other things that the whole complex process. And so the exciting thing is with Big Network, that's a problem that you're making very, very, very easy to solve. And so with Big Network and this flat network that it helps produce, does it offer encryption as part of that?

Tom Daly: [00:35:07] It does, Yeah. So we have a key distribution system, private public key distribution system. It's all based on AES 256 bit encryption. It forms a full mesh of tunnels automatically. You don't have to go form that full mesh of tunnels like you do with IP SEC, which ends up getting into like the tunnel scaling problem, right? So you end up getting this non geometric or more than geometric sequence scaling scheme that you have to get into when you're manually configuring the tunnels yourself. And yeah, everything is secured AES256.

Jake Warner: [00:35:44] Excellent. Excellent. And so in terms of is so with with with with the devices, the software that you have, is that something that the public can now try out? Do you have a beta period open? Like if someone is listening to a podcast and they want to try it out, how could they?

Tom Daly: [00:36:01] Yeah. Anyone can go to our website, which is just Bignetwork.com. There is a sign-up button on the top, right? And you can get a free trial for your devices, meaning your endpoints. So your Mac Windows, Linux, iOS, Android box. We do ask folks for credit card number up front. If you sign up for our core plan, we won't charge your credit card, but we like to have the credit card on file as a anti-abuse mechanism just to weed out some fraudulent signups. And then we also have a friends of Big Network early access program to Edge Light. And so if you are interested in trying out the hardware box, shoot us a tweet on Twitter. Get big network, send us a message on LinkedIn or shoot an email to sales@bignetwork.com. And if you're a network engineer, software developer, or have a networking background, we will happily send you a edge light and trial credits for a year to get a sense of of the hardware as well. I think you have one, Jake, right?

Jake Warner: [00:37:11] I do. I do have one. And I remember when we were going through the early days of setting it up and you're like, you want to see something neat? And from my phone, I pulled up the from my phone over the L2 network that I'm sorry. Yeah, over the yeah, the L2 network that I created. From my phone after disabling wifi, but still having the application running. I was able to access my link, my router's control interface. It was pretty neat knowing that I could if I ever needed to.

Tom Daly: [00:37:49] You're on a local LAN.

Jake Warner: [00:37:50] Right. Exactly. And being able to do that from anywhere is really neat because who knows? Like when is when is the nest camera going to not connect anymore and I need to reconnect it. Or, you know, who knows? And it's just really nice to be able to have that functionality or that flexibility of being able to do that anywhere.

Tom Daly: [00:38:07] And what's interesting in this for for someone at home, a geek like us, right? I mean, we we're learning new use cases every day. And one of the most interesting ones that we just came across, a friend of ours brought this to us. You know, there's a lot of devices that have some sort of cloud for user control, right? Like, you know, your Nest thermostat uses Google Cloud and exposes a beautiful app so you can control your thermostat. But there's some out there that we found. And one of the most interesting ones is the Roku TV streamer. So it is cloud managed. It can be controlled over the cloud, but there's one critical button that you can't press remotely and you have to be on your local network to do it, which is to enable guest mode. And so if you're running an Airbnb, for example, this was my friend's use case. He's running an Airbnb if he forgets to enable guest mode before he leaves. The only thing he can do is drive back to the place, get on the local wifi network and enable guest mode for Roku. And we were like, No, no, no, no. Like, let's show you a Big Network. Let's show you how you get tunneled L2 access in with absolutely no changes to your firewall, no dynamic DNS, no port forwarding. It's super, super easy and like just turned on roku's guest mode like that. Remotely.

Jake Warner: [00:39:28] I mean, it's yeah, it's, there's so many different things that you could do with it, even from even from like if I had a file server that I was running locally and I was like, Hey, instead of exposing this to the Internet in a traditional sense because I wouldn't need to access it for somewhere else, just having that flat, that flat network that just makes that seamless is just there's so many opportunities there. And then again, from a Cycle perspective, I'm just sitting here. I'm like, my mind's already racing. Like, you know, what are the other things that we can we can help solve problems with, especially with the IAL now being live. We've had more and more companies coming to us recently saying, Hey, we have this on premises infrastructure. We want to gain the capabilities of a traditional cloud, but we still want to own this infrastructure. Actually, that's a question I want to pose to you. So with Cycle, we actually have more bare metal running on the platform today than virtual machines, and we're seeing more and more companies switch back to bare metal, not necessarily throwing it in their closet, but switching back to bare metal. Is that something that you've been seeing in the industry as well?

Tom Daly: [00:40:26] I've been. I've been seeing it and living it. Right. I mean, you know, you take an infrastructure like Dyn or you take an infrastructure like Fastly and even to a certain extent, an infrastructure like Big. What we're operating on the background and there are a whole bunch of elements of our system in stack that make a ton of sense, put it in a cloud provider, like just do it, don't try to run it yourself. But if you think about a 100 terabits per second caching CDN network. From an economic standpoint, that does not make sense to put in any cloud provider. I mean, you will just you'll just get eaten alive on bandwidth egress charges. And when you're running these large global infrastructures, you want to be in control of things like your network and your transit relationships and your peering interconnection relationships and what you might be doing with something like a private backbone. And you give up some of that control of the network when you're in a cloud provider. So I think that everything sort of comes around, right, in our industry. It was mainframes and then it was client server.

Tom Daly: [00:41:43] And if you think of the cloud, it's kind of mainframe you again. I mean, I think we've seen a bunch of stuff move to the cloud and there's a bunch of workloads that are going to stay there because they just make sense to be up in a cloud provider. But there's going to be workloads that come back out. Right. And it's those workloads that are specific and niche. They have specific hardware requirements. They have specific topology or network requirements. And, you know. It's my belief that the cloud providers from a network standpoint are never going to play nicely with one another because that's their moat. And so are a lot of their moat is the network. And so let's figure out how to how to abstract that mode and make it easy to get over. Give everybody a drawbridge over the moat, no matter where they're trying to place their infrastructure. Yeah. And so I think that hybrid cloud is up and coming and it's going to be here to stay for a long time. And it's probably it's probably for a good while the right way to do everything.

Jake Warner: [00:42:46] And I fully agree, like I think I mentioned to you before, but more than 60% of cycles customers are using multi cloud or hybrid cloud today in some sense. Right. And it's really interesting because, you know, you hit on the point about performance. But the other thing that we're seeing is, is at least with platforms like cycle is the increased density on these bare metal servers, right? Like if you go over to AWSs and you say, Hey, I need a server with 20 gigabytes of RAM or whatever, you might get the server that's 20 gigabytes of RAM, but you might get like five VCPUs in it, right? So you're getting maybe 1.15 or 1.135 Physical cores and you're paying $500 or $600 a month for where there's some of these infrastructure and service providers now that are offering bare metal where I mean, take Vulture, for example, with Vulture, they have a server right now that is 32 gigs of RAM. I believe it's eight physical cores for $122 a month. Right. And I mean, you go over to AWS and I mean, for those types of resources in a VM, you're going to pay ten x the price. And so it's one of those things that I think people have been afraid of bare metal after adopting the cloud because they're so used to like, Ooh, I can scale up and scale down and everything. But, you know, as time goes on, we're seeing more and more companies starting to, you know, adopt a lot of this bare metal, even if it's in infrastructure as a service like provider, just because there's again, if you can get that increased density, you can get that performance. As you mentioned, there's so many wins that you can get there that it's like.

Jake Warner: [00:44:15] I don't know. I, a few years ago cycle as a company, we were running most of our stuff on VMs in a traditional cloud. But now today we run more in bare metal still in these cloud providers. Or maybe not. Maybe cloud provider is the wrong term. But these infrastructure-as-a-service providers, we're running more bare metal in them than we are VMs now and then our customers are the same. So it's just really interesting seeing that progression had happened. And one of the things that you and I have talked about previously, and I think I mentioned it on a podcast before, but we always see that pendulum, right, where you always switching between simplicity and customizability and it always is almost about every seven years it's flipping back and forth, right? And even right now within the within the container orchestration space you have for the last five, six years, you've had all these people switching over to Kubernetes because they want that customizability. But through a lot of the conversations we're having, we're seeing that pendulum starting to swing back the other way towards simplicity. People are saying, I don't need to be able to customize a thousand different things. I just need to be able to customize these things. And I think that kind of aligning that with bare metal or even with what you're building with big network, it's like, like, yes, I want that Customizability. But more important is I want that simplicity. I want people to just plug in these devices and go and it's just one of these things that it's just really interesting to see happen within the industry.

Tom Daly: [00:45:38] 100%. I mean, I think, you know, bare metal for a period of time as cloud was adopting, like bare metal meant the traditional on premise data center or the traditional co-location site. And it was a dirty word, right? Like we didn't have DevOps, we didn't have SRE, we had developers and sysadmins, right. And developers wanted to induce change into this, into the system and sysadmins wanted to prevent change in the system because they didn't want things breaking. And you know, here comes the cloud, which empowers developers to like swipe a credit card and get an infrastructure that they're in complete control of without, you know, all the I mean, the places we started where it was like, you want a server like, you know, get out a floppy disk or get out the CD rom or get out a USB drive, like you're walking downstairs into the cold air conditioned data hall, Right. And there's this whole generation of developers that, for better or for worse, like the start of a server for them is some API command or some CLI tool like I was I, I am not a developer, let's be clear. But like for many of my years, my starting point for a new server was something like knife, EC2, create, right? That's how I got a server for a period of time. And that became so de facto so understood that we were going to get our infrastructure via API and command line that I think the concept and you just can't really do that with bare metal, right? You can't. But we've had so many infrastructure as a service providers or bare metal as a service providers come up and provide that level of abstraction to the actual hardware now where bare metal is very much real, capable, extremely customizable and powerful.

Jake Warner: [00:47:28] Yeah. And as I mentioned, it's nice that you're seeing that in the industry or not just seeing it, but living it as well. Because for, for a while now we've been seeing that and it's one of those things where when I bring that up to people, there's a lot of people that are like, really? Like you're not just going to throw everything at EC2 instances. I'm like, No, like there's a better, more efficient way of doing these things. So it is interesting that you're encountering that as well.

Tom Daly: [00:47:55] But I think I think there's a time I think there's a time and a place for everything. If you're, you know, if you're going to build some social app, right? And you're starting out as a company, you got your first, you know, 3 to $5 Million of series seed funding, and you're going to put that to work and build your app. Like, I think you're crazy not to go build it inside one of the cloud providers, right? Because it just gives you so much foundational infrastructure so fast. The problem comes when you're doing massive levels of scaling inside the cloud and you can't see well, you can see, but you can't control the inefficiencies that end up existing. And that's where, you know, adopting or at least having a plan in an architecture that allows a hybridized approach early on is important. Yeah.

Jake Warner: [00:48:43] I completely agree. And that's actually one of the that's one of the stories that we've been telling to a lot of these companies, even in the early days, is, hey, you know, if you adopt a platform like Cycle, you know, we make it so easy to build these hybrid environments. Even if you are going to start with VMs, we're going to make that path easier to bare metal should you ever need it, right? And so so again, being able to eventually, someday pair that with Big Network is something that gets me really excited. But as we get closer to wrapping up the podcast today, as I mentioned in the introduction, that, you know, you're an angel investor. You're an investor in Cycle. But just regardless of the Cycle component there, curious of sharing your perspective on, you know, startup investing and things like that with our audience. We have a lot of startup founders that our customers of Cycle that listen to this podcast, etc. and I think they'd be really interested in your thoughts on on what companies you look for. What makes a company good to you? Things like that.

Tom Daly: [00:49:44] Yeah, I'd be happy to share. So, you know, I believe in a strong entrepreneurial ecosystem. I think entrepreneurs and small businesses are the data says they're the foundation of our country in many ways. But yeah, so in terms of what I'm looking at, what I look for and the theory behind investments, you know, first of all, it needs to be in an area I know something about, right? So, you know, I know Internet infrastructure, I know some things about software. You know, I work at various levels of the Internet infrastructure stack. So I always invest in areas that I have a healthy level of familiarity with. So I'm not going to get involved, say, with, as an example, a biotech startup. It's just that's that's not my jam. That's actually my wife's jam. She understands that way better than I do. So I stay away from areas that I don't have some preexisting knowledge of. But from there, you know, then I'm looking at a couple of different things. You know, I'm looking at how does it lever into my background, how does it lever into the founders background? Have they lived the problem that they're trying to solve in some way? So, you know, a lot of people talk about product market fit. There's this pre stage to that which I would call founder market fit and I think that industry or that term has been making its way around the the venture capital community.

Tom Daly: [00:51:08] You are looking for founders who deeply understand, empathize, sympathize and can solve a problem that maybe they live themselves or or for folks that they have been in service to over the years have had to solve the problem for in the past. Some of my favorite founder profiles are folks who are in the professional services world, and they have become an expert in their field, whether it be. I'll give you an example. Will Mitchell is the founder and CEO of a startup called Vetro Fiber Map. He is a GIS expert that specialized in providing GIS consultancy for people building fiber to the home networks. And it turns out when you're building fiber to the home, you need a lot of GIS systems to support that, to figure out where you should deploy, how you should structure your fiber cable, whether it's underground or on a pole, and then use a lot of GIS data to operate your network. You know, when there's a fiber cut, there's these things called OTDRs, optical time domain reflectomoters that will help you locate in distance where a fiber break is, right? Well, if you're standing at the data center headend and it says, hey, the fiber break is 30 kilometers out, you know, if you don't have good GIS data, guess what you do? You jump in a truck and you start driving along until you find like the car into the telephone pole, sadly.

Tom Daly: [00:52:40] And you know, when when unfortunately, it's a drunk hunter that has decided to take shots with their rifle at your fiber cable because it's fun. It's pretty hard to see that from a vehicle. So if you have good GIS data, you can look that up and say, hey, based upon the way that the fiber is spinning in the slack loops that are in the aerial plant, we know that this is going to be between pole 111 and pole 112. Let's go find the find the break in the fiber. So Will has an amazing background in GIS consulting. And he decided like, hey, I seem to be doing the same thing for my clients over and over and over again. I should just make that into some software. And that's essentially how a company like Vetro Fiber Map was born. And I love seeing, you know, companies that do those sorts of things. And I love investing in founders that have those areas, those expertise because they've lived the problem. They deeply understand all the nuances. And in some ways, like the product roadmap is just in their head. They just can see where they want to go on a multi year horizon, a very, very natural fashion.

Jake Warner: [00:53:46] And that makes complete sense. I appreciate you sharing that. But I have one final question. So with the companies that you're invested in, So you shared a story with me a few months ago and ,You were getting ready. You were one of the startups that you had invested in, was getting ready to dig a tunnel or bury some sort of line underneath a river. Can you can you dive into that?

Tom Daly: [00:54:24] Yeah, happily so. A few years ago now, I got involved with a gentleman by the name of Josh Snowhorn. He is the founder and CEO of a company called Quantum Loophole. Quantum Loophole seeks to build out the largest scale, environmentally friendly data center campuses all over the world. And our first project is happening in Frederick County, Maryland. Frederick County, Maryland is about 30 miles north of Ashburn, Virginia. For the geeks on the call, everyone knows that Ashburn is effectively like the center of the North American Internet. There's tons of submarine cables that come in and out of Ashburn, all of the major cloud providers there. Not to pick on our friends at AWS, but any time there's a US East one event that is all happening inside Ashburn and the challenge in Ashburn is Ashburn is full. You know, when I was going down to Ashburn in the early two thousand like it was farmland, and now it's all data centers. So they're running out of land. And we've also heard from the local utility provider, Dominion Energy. They really have very limited capacity to supply additional power to the Ashburn market until 2025. So the hyperscale data center community is very challenged in Ashburn right now. And so Josh and his team at Quantum had the foresight of this coming and decided that they were going to start building some of the world's largest data center campuses.

Tom Daly: [00:56:03] So we identified a parcel. The team identified a parcel of 2200 acres of land in Frederick County, Maryland, with a significant amount of gigawatt scale power available and in such a way that reduces the need for non environmentally friendly power sources, i.e. diesel generators. The problem in Frederick was that there just isn't that much fiber there. You know, Ashburn is 30, 30 miles away and we just did not see a way to get the traditional means for middle mile connectivity up into Frederick. So the good news is the team at Quantum is very diverse. And not only do they understand dirt and power and water and sewer and buildings and powered shells, but they also understand fiber. And so Josh and his team set out to build one of the world's largest middle mile networks, 235,000 strands of fiber from Loudoun County, Virginia, up to Frederick County, Maryland. But with one big problem. And the big problem is, is that all the interstate bridges across the Potomac River that you would normally hang fiber from are completely full. So the Department of Transportation in Virginia, Department of Transportation in Maryland say, nope, no more attached attachments for fiber infrastructure to our bridges. There's too much weight. And in recent history, there hasn't been a significant fiber crossing across the Potomac for infrastructure in many years.

Tom Daly: [00:57:48] So we couldn't go over it. So we decided to go under it. And you can go check out our LinkedIn page. Our drilling machines are hard at work. These are directional drilling machines that are drilling a 12 inch borehole under the river, about 2000 feet long, popping up to the other side. So it's the machine where you can control how it goes under the river. And once we get that 12 inch bore through, we then attach another device called a back reamer, and we pull that back from a pilot to 12 inches. We pull it back to 24 inches, and eventually we open up that borehole to 36 inches, at which point we can then slide a 36 inch HDPE sleeve. So basically think of like high pressure natural gas pipe that we slide under the river and then we can install roughly 32 inch and a half HDPE conduits. And through that you start pulling your actual multi thousand count fiber cables. So system design capacity of 235,000 fibers. It's taking three different river crossings to across the Potomac and one across a little river called the Monicasy. And that will extend the Loudon County Internet ecosystem with extreme scale to 30 miles north to Frederick County, Maryland.

Jake Warner: [00:59:19] One. I mean, that is incredibly fascinating. But number two is I can't imagine all of the permits that needed to go through to be able to to go under a river like that, unless it wasn't. I mean, I guess you you would definitely know more than me, but was that like, how long did it take for them to get that sort of permission to be able to do that.

Tom Daly: [00:59:45] Order of years is a reasonable approximation for getting this project off.

Jake Warner: [00:59:49] Yeah, I could imagine.

Tom Daly: [00:59:50] There is a there is a ton of permitting that goes into that. Right. I mean, and one of the things that I appreciate about the Quantum team is, you know, they're doing it. They're doing it right and they're doing it environmentally friendly ways to. So I think that there's a real win for the industry as a whole there. And they're going to set an example in a high bar for people that come behind them.

Jake Warner: [01:00:12] Excellent. And so that specific area is the only project that they're working on right now. Or is it? As much as you can.

Tom Daly: [01:00:20] It's 2200 acres. We're kind of busy, my friend.

Jake Warner: [01:00:23] I know. I mean, the specifically the, Well, so maybe that's a point of confusion for me is so is quantum loophole mainly focused on developing everything or was or is there strength being able to get the fiber to where it needs to be?

Tom Daly: [01:00:42] Yeah. So Quantum really provides a couple of different things, right? Like we don't actually provide data centers, but we provide all the things that a data center needs, right? So we provide land, we provide electricity, we provide fiber connectivity, we provide water and we provide sewerage, and then we provide all the public facilities, roads, bridges, traffic studies, all the things that you need to actually have a massive scale data center campus exist, right? Because we need these things to like fit into our communities. Right? And it's not just like plop it in and you're done. It has to be very, very well planned as to how it fits into the geography around it.

Jake Warner: [01:01:21] Yeah, and that makes a lot more sense. Why for the disconnect of again, was it was it just the fiber but knowing that they're handling so many of these other things? Yeah, I get that that that single project is you know I'm guessing what a decade plus effort.

Tom Daly: [01:01:37] Josh and Josh and his team have been very, very busy and they continue to be very busy, but they're stellar people. They got some great stuff going on.

Jake Warner: [01:01:45] Excellent, Excellent. Well, Tom, thanks for taking the time to be on our podcast. Really enjoyed the conversation. For anyone who wants to learn more or connect with you, what's the best way for for people to find you on the Internet?

Tom Daly: [01:02:00] So let's see, LinkedIn is probably my favorite. So just search for Thomas Daily on LinkedIn. Hit up Big Network on Twitter at the handle, @getbignetwork or pop on to our website and send us a note- Bignetwork.com.

Jake Warner: [01:02:16] Excellent. Excellent. Well, thank you again, Tom. It was great.

Tom Daly: [01:02:21] Thank you for having me, Jake. It was great to talk to you and talk to your audience.

Jake Warner: [01:02:25] For anyone who's interested in finding more of our podcasts. All of these podcasts are hosted on Cycle.io, our website. You can also find them on YouTube. You can find them on Spotify, Apple Podcasts, and I believe Google podcasts as well. We've had about a four month window where we had recently paused on podcasts but starting with this podcast. The goal is to ramp podcasts back up and continue them monthly. So thanks for listening and check us out on Cycle.io