Peering, edge computing, and community with Grant Kirkwood

Media Thumbnail
  • 0.5
  • 1
  • 1.25
  • 1.5
  • 1.75
  • 2
This is a podcast episode titled, Peering, edge computing, and community with Grant Kirkwood. The summary for this episode is: <p>Chief Technology Officer and Co-founder of Unitas Global, Grant Kirkwood, joins Network AF to discuss motivations for starting the company and where they're at currently. Avi and Grant talk about what it is like to be a service provider and a solution provider (MSP) in one, and how it plays into what Avi calls the APIfication of networks and IT strategy. Topics throughout the conversation include: </p><ul><li>Non-software enterprise companies and whether or not they'll adopt more interest in peering</li><li>Donut Peering and creating better interconnection</li><li>Building reliable and resilient networks for edge computing</li><li>Unitas' growth and how Grant thinks about building networking skills internally</li><li>Looking for people to join the team who are curious</li></ul>
Episode overview and intro to Grant Kirkwood
01:59 MIN
Why Grant wanted to start Unitas
02:28 MIN
What it has been like having a service provider and solution provider bundled together in a market where people are less opinionated due to on-demand consumable cloud services
03:33 MIN
The APIfication of networks and IT strategy
04:26 MIN
Grant's thoughts on whether or not companies in non-software enterprises will adopt more interest in peering
05:08 MIN
Donut Peering, "packet destiny," and creating better interconnection
07:15 MIN
Building reliable and resilient networks with the concept of edge computing latency
08:42 MIN
Unitas' growth and Grant's thoughts on career progression with networking skills at the company
04:36 MIN
Looking for people with curiosity, and how that trait can show up during interviews
04:27 MIN
Cosmology, physics, and system-level thinking
04:41 MIN

Avi Freedman: Hi, I'm Avi Freedman and welcome to Network AF. On this week's episode, I'm talking with Grant Kirkwood CEO of Unitas Global. We're talking about cloud, how it's built, what's behind it, peering and how the cloud gets to end users, edge connectivity, and the state of the art, and even more the state of the business about how people are getting connected, and then community, and the community side of networking and how to get in, you can get into a career path in networking companies. Hi, and welcome to Network AF. This week, I'm talking to my friend Grant Kirkwood, who's built a number of successful businesses and run them from into the nineties and the dawn of the internet to leading edge cloud and connectivity technology. We're going to be talking about cloud, about peering, about the edge, about how people get connected, and also about community and how we get people into networking and how we grow our diversity into the next generation. Grant, if you could give us a quick intro on yourself?

Grant Kirkwood: Sure. Great to be with you Avi. And yeah, so I've been in and around internet infrastructure for basically my whole career. Started my first company in 1996 when I had a lot less gray hair than I do today. But yeah, I grew up in the early bulletin board systems and very slow modems and just fell in love with it. And so yeah, Unitas is my third company, but have started a couple others before. Really just like the plumbing and the infrastructure that most people don't really see as part of the internet and cloud and the physical manifestation. So it's always been my passion. So anyway, glad to be here.

Avi Freedman: Thank you. And thank you for helping the world stay connected. I love that I get to work with folks that make the infrastructure and then the applications that sit on top that people use it. It's pretty cool. So Unitas, as you mentioned, not your first company, what got you to want to start Unitas and how do you think about where your focus is?

Grant Kirkwood: Well, it's funny because Unitas, being my third company, I wanted to do something a little different. I'd been in the internet infrastructure space for my whole career, and so I thought," Well, let's combine all the IT resources that enterprises consume and provide that as a fully baked managed service that includes everything. And that's literally what Unitas means in Latin. At least we say it does. But the whole idea is that we could do that on a global basis. And what's happened in the time since we started providing those services to today, the need for very custom, bespoke IT solutions has decreased with the prevalence of AWS, GCP, Azure, et cetera, but the need for more sophisticated networking solutions has increased. So, come full circle, back to doing what I've spent the early days of my career doing, which is fun and exciting, although the world's changed a lot in those years, at least in some ways, not in other ways. Still fiber, still data centers, small conduit in the ground that makes it all work.

Avi Freedman: It's pretty amazing really that now actually the world has simplified in that, yeah, there's optical underneath, but from a IP basis, it's not like we're hooking up Sonnet and understanding all that stuff that, although there are good things about that from a telemetry perspective and debugging versus ethernet, but it is pretty amazing that ethernet is effectively the same thing as when we were doing vampire taps or maybe you escaped vampire taps, but just moving from hubs to switches and all that and it still works. And you think of the whole internet as actually running as a serial protocol, one bit at a time underneath, you're just awed at how fast computers are, versus as you mentioned, the era when people got into it. So, it's interesting. I hadn't thought about that really, but I learned more when we started Kentik about... In the US people think about internet connectivity as something I go get to the internet provider and then as you said, IT stack I go get from others. I know that in AsiaPac and Europe, we have customers where they are much deeper into the IT stack for their companies. Now, some of them are thinking about, well, we don't want it all to be AWS. So we need to do, as you said, bespoke networking and also other solutions. Is that, I want to say a business cultural problem for you, when you took this approach in the US, that people weren't expecting to have a service provider and a solution provider, or MSP all in one, or has that been an advantage?

Grant Kirkwood: I think it was initially, because in the early days, people focused on what type of storage array or server, or what kind of software is running around and they're very opinionated about those things. And that was fairly globally, I would say. And people have become much less opinionated, right. So it used to be, I need to have a rack that has this type of thing in it. And I think the rise of on- demand, consumable cloud services has negated the need for that level of specificity. On the other hand, in the early days, the, how you connect to that thing was almost an afterthought because the consumption was driven by the IT buyer in the organization, as opposed to someone with a broader remit. And then the connectivity or telco buyer bought their pieces and they're almost quite distinct. And so it was an afterthought. And what's happened is that's shifted too. Now you find that the consumers of those same IT services in cloud form are now also consuming network services and they have the expectation that they can consume those services the same way that they're used to spitting out virtual machines. And the business hasn't quite caught up to that industry- wide, which I think is a great opportunity for our business and others that have taken a more modern network- as- a- service type approach to consuming those at that capacity.

Avi Freedman: So really, if I think back to a dinner that I had when I was at Service Central and we had a vendor in, and we had, I think, 10 people from the vendor, because we had the account rep, the account sales engineer, the storage sales engineer, and then one rep and sales engineer from each of the five things that they had two inaudible and an object store thing that didn't work and an object thing for a thing that was going to work at some point and a filer and trying to talk about all this, and for all of them the network was not 10 megabits, but gigabit or something. And I was like, you do understand that this isn't the dawn of big data. You do understand that that would be like sucking the ocean through a straw to put half a petabyte on the other side of it, even a 10 gig connection, how long would it take to do something? And even though cloud is all these things, it was a very different focus at, I would say, at the dawn of it. But it almost sounds like the APIfication, network as a service, the enterprises want to consume that way and to talk about cloud. So I've always thought of cloud as other people's network storage and compute. Now everything is infrastructure as code, and so cloud is effectively, in some sense, it's just all these technologies we know with weird names consumed via API, but the enterprise sometimes use it at a higher level and so how do you see your customers thinking about cloud? Do they understand that network? Are they getting bit by the fact that they don't take network into account or do they understand that networking is a critical part of that? How do they view that strategically and then as they bring it into their operations?

Grant Kirkwood: Yeah. There are some enterprises that we see... Let me give a little context first. Our customers tend to be either large regional or multinational enterprise. They're typically locations and different continents and hundreds or thousands of employees. And so what I would say is there is a subset that is very focused on network and connectivity as an underlay to their broader IT strategy, but they're in the minority and it's usually because they've had a connectivity- focused practice as part of their business. So as an example, we have a customer, that's a healthcare provider, they provide software for hospitals and health care facilities. And as part of that in their traditional model, they actually procured last mile circuits into those healthcare facilities to provide the delivery of their software. And so they had what looked and felt very much like a telco within their business. But as I said, that's very much in the minority. For the most part, we more commonly see enterprises being bit by the fact that they didn't consider the needs of the network as they looked at what their IT strategy is. And especially in the last two years, in particular, we've seen this in a big way where all of a sudden we have this massive workforce that shifted to work from home, all of a sudden jumped on zoom and Teams and WebEx and all these other things and turned on video for the first time and got very used to that mode of working. And then fast forward to the last six, nine months, all of a sudden you got a bunch of people back at the office, still some people at home, let's say it's hybrid half, half, whatever and finding out that all of a sudden, all that office doesn't have enough internet bandwidth in the office to support the number of remote video users that are now collaborating from home and very quickly scrambling to get more internet capacity.

Avi Freedman: Interesting.

Grant Kirkwood: And that kind of scenario is very commonly playing out. I think also the shift away from traditional NPLS to SDWAN as a quasi replacement, alternative same thing there. We've seen a lot of enterprises that just go and buy whatever local connectivity they have from whatever local provider in that particular market exist, and assuming that it's all just going to magically work and there's going to be this good, consistent experience that looks and feels like their traditional NPLS. And then quickly find out the regional provider in South Africa that they bought from doesn't peer with the tier one ISB that they have in North America and so therefore that underlays is a bad experience and that impacts what they have inaudible the top. So, yeah, I think probably 70, 80% of the time, more commonly, we find the enterprises are saying, oops, as it relates to the network.

Avi Freedman: So I just want to poke at that. When COVID started, we had a lot of customers that were saying, Hey, how can we look at what will happen if we unleash the kraken and do split tunnel instead of bringing everything back through our corporate network. But that was as things were starting to go video, as you point out. So you're seeing the reverse effect of people being trained on new patterns and then coming back, even if it's less frequently, but to offices which wind up just having wound up under provision, just because the usage patterns had changed. That's interesting. We probably could scry in our data and see that, but it wasn't something that I was aware of was going on.

Grant Kirkwood: Oh, very much so. And then you've got the enterprises that did the traditional, centralized internet decryption and inspection stack. One crosstalk

Avi Freedman: But that broke two years ago. That broke two years ago when everyone was trying to come in.

Grant Kirkwood: Broke in a day. Exactly. Right. Literally overnight, they had just enough internet for the office when everybody was meeting in the same room, went home, all of that internet for, I think in that case 40,000 employees, all of a sudden was going over VPNs back to their stack. And it broke, it broken a day. They had to break out that local internet by policy literally overnight or that business couldn't have continued to operate.

Avi Freedman: Interesting. So it's been interesting as we work with service providers, they certainly understand peering and interconnection. Digital enterprise that really started before or after sneakernet was ever a thing where the business is the application that run on the network. They understand the value that, even just from a performance optimization... In the nineties peering was like, ah, I'll save a lot of money. I don't have to pay X as much, but right now we see it more from the software companies think about it as control and control over performance and the destiny of my packets. At least I have the food when there's a problem to go do something rather than calling someone up or emailing them and hoping they'll do something. We've definitely seen a lot of enterprise that are more traditional that come at it from a perspective, oh, I have a inaudible here, as long as CDN will get to me, it'll be fine. Do you think that the average, I'll say mainline enterprise, that's becoming a software company, but did not start that way, are they becoming more interested in peering? Or are they more looking for someone to help them solve the problems that might happen, when, as you said, I'd go back to the office and there's congestion or I shifted from on- prem to a bunch of SaaS services, which take a lot of bandwidth. How do you see that going in the non- software company enterprise?

Grant Kirkwood: So I would say they don't know to be interested in how that interconnection happens. So they're very interested in making it work well and very interested in getting content or applications to users as quickly and efficiently as possible. And I think there's probably a defacto assumption that, well put it on CDN and that's going to happen. What they don't understand is why the very aggressively peer- with- any- eyeball kind of network approach that the CDNs have, contributes to that. They don't understand that piece typically. And in the enterprise space we do a lot of education about what is a tier one ISP and what is a tier two ISP, and does tier one mean better or not? And the typical enterprise, their basis of knowledge is the marketing slicks that are put out by the large global carriers that basically tell the story of bigger is better. And then when you explain that, well, actually the reason that CDNs are so efficient at what they do is because they take the exact opposite approach to interconnection with other networks. And so there's this education process, and when we literally draw the picture of the donut and the tier ones and the tier twos, and by the way, all this stuff out here is what you actually are connecting to, that light bulb goes off, but it takes education. When we have that conversation with the enterprise, they get it and it becomes important. But to your point, they don't know to be interested or pay attention to that without that level of education in terms of how the internet actually works.

Avi Freedman: Yeah. It's been interesting to see one of our customer Netskope does a lot of aggressive explanation really about not only why it's good to be using someone that's broadly connected and can be better than, again, the CDN argument, better than just throwing up a couple of connections and doing it themselves. But it does seem like in the world of APIfication the default is, well, the packets will... whatever. Packets? What? Or they'll just magically happen. When I was at Net Access we took this approach because I wanted to... The reality was... Little, dirty secrets. Actually did peering because it was$ 800 a megabit or$ 50 a megabit. I'd rather have$ 50 a megabit. So the actual, it started with cost in the nineties, for sure. And probably a fair bit of ego, I am doing what the people that build the internet do. And just the engineering, I want to understand the internet, which is this cool thing. But it was true. We noticed when we started doing performance tests that, as you say, there was this bifurcation and we tried to make it about an end, like yeah, yeah, buy from someone who's in the club, but the people in the club don't directly connect to this host of other people. So you talked about donut. I know what that means, but maybe for people listening that don't, would you like to explain? What is the donut appearing peer? What is this bifurcation that we're talking about here?

Grant Kirkwood: Well, besides chocolate and sprinkles.

Avi Freedman: Yes, exactly. Besides that.

Grant Kirkwood: The whole concept of donut peering is that you really, if you look at the internet as a whole, there's, I think 73, 74, 000 ASNs on it today, somewhere in that neighborhood, although it keeps growing. There's a very small number, double digits of those, which are effectively tier one ISPs that have settlement- free peering with 100% of the routing table. And I like to use the analogy of fight club. You don't talk about it, and once you're in it, you don't let anybody else in it. And then you've got somewhere on the order of 5 to 6, 000 ASNs that are the classic tier two networks, half of those being content, half of those being eyeballs. And if you look at the traffic on the internet as a whole, something like 75% of traffic transits, originates, or terminates between those tier two networks. So that, classic analogy I always like to use is, somebody at home on their cable mode, watching something on Netflix. Some of the very largest cable providers are very large, but still tier two networks that buy transit from tier one's as is Netflix. And that represents a huge chunk of traffic. Microsoft, Amazon, Google, Facebook, go down the list and then line up all the eyeball networks and they very much fall into that same category. And so if you think about the donut with the 5, 000 or so, ASNs that sit out there in that tier two ring and the small number in the middle, the whole idea is that if you're a tier two eyeball network and you go around the donut and peer directly with a tier two content network, you're effectively creating a bypass ring around the tier ones in the middle. So that's why we call it donut peering. But it does demonstrably improve performance, fewer hops, lower latency, better throughput on a whole.

Avi Freedman: Less transfer congestion. And now we have the tourists, which is really, if you had the plane, which is the overlay networks. We have customer that feed us their DNS, and we can see what sites they're going to and 90 to 95% of their traffic is these meta networks, as you said Netflix or CDN like inaudible or whatever. It is pretty fun. And at Kentik we have BGP analytics, so we can see this and we classify edges and who's getting transit and all that. But my favorite, speaking to fight club, my favorite informal algorithm for looking at tier one versus tier two, is the best way to find out who's tier one is to ask every network who are the tier ones, they will all name the actual tier ones and themself. And then you remove the identity mapping, and then you're left with the people that really are. But, yeah, from a strategy perspective, I really believe, I'm not just trying to be fair to our tier one and tier two customers that having some connectivity from the donut hole and some from the donut is what you want to do. And I'm hopeful that as people think about taking their packet destiny into their hands and connecting to cloud, starting with cloud, whether you do it yourself, or whether you use a network as a service or your service provider to do it, I'm hopeful that can get people into it. Because, well, I mean, we're seeing a little bit right now of at least hot topic around splinter netting, and what's going on with Russia and Ukraine from an infrastructure perspective and Russia from a center perspective, if you don't have enough connectivity. And so far, I've just been thankful the connectivity's rich enough that when the physical infrastructure is working it's mostly working.

Grant Kirkwood: Just on the previous note for a second, totally agree with you. We talk to our customers all the time about the best way to get the most diverse routing table is with a tier one and a tier two.

Avi Freedman: Interesting. Even though you're a service provider, you'll recommend. That's why we did at Net Access because it was a battle we weren't going to win saying we were better than level three or Verizon or whatever. And it was based on technical fact, right? This is the best way to control your destiny.

Grant Kirkwood: Yeah. If you MultiON between two tier ones, you're not creating a lot of diversity in your routing table. If you connect to a traditional tier one in a very aggressively peered tier two you're creating a lot of diversity on your writing table and that gives you optionality. And to your point about what's going on with Russian and Ukraine, I think the same thing applies a lot there too. You look at some of the Russian ISPs that sit on the peering exchanges throughout Europe, LINX, inaudible et cetera, they get a lot of their connectivity through there. I know that some have been disconnected as a result of the sanctions, but something I've been talking with people about recently is the unintended effects of sanctions that it effectively completes the, call it the great wall or iron curtain or whatever you want to call it, of internet censorship. Whereas everything that's been done within Russia to this point has really been layer four and above. And we've seen that VPN usage in Russia grew something like 8000%, just in the first week of the invasion. And so there is a sizable percentage of the population that's using APM technology to get to western news sources and whatnot. But I worry that if you cut off access to the rest of the routing table completely by turning down transit and peering with Russian ISPs, then that no longer becomes an option. Then you literally do have a splintered internet. Right now, from a layer three perspective, there's still inaudible

Avi Freedman: There's been discussion about it, but definitely some connectivity has changed, but it hasn't been a bifurcation. It hasn't been cut off. The only side that's been cut off, cut off is just when fizzle infrastructure is destroyed by the tragedy. When the last cell tower goes down there's no connectivity. So yes, hopefully putting the infrastructure back together will work and be pretty quick. And the good news is the technologies that we've had since the nineties work pretty well for this and are working to keep things going, even as we can definitely see impact as infrastructure is being attacked physically. It's sad. We think about infrastructure being attacked for decades as a logical attack and now we actually have it being physically attacked, but it is mostly staying up and fingers crossed on that. So peering, the point is to get between networks, which can be eyeball or content, or both, maybe that's a naive view. That was the nineties model. I think of edge. There's been a lot of, oh, we'll put trailers in parking lots, the servers can talk to each other. And before COVID, at the end of 2019, I had just been to a number of customers where they actually had a little data center in a closet and I had this realization, it's like, wait, this is the actual edge. The actual edge is not applications looking to run today, it's not this or flying cars or whatever. But the other side of the edge, which became critical with COVID is at home. You said to me something which is that you had a conversation where people were talking about the fact that the business model, that hasn't changed. I'll ask you what was that insight? What was that conversation that you were in? That was about access for home and business?

Grant Kirkwood: Yeah. I would agree. I think the concept of edge computing is still the on- prem server in a closet model and that's still in place in a lot of businesses. Although I think decreasing, generally speaking, at least certainly as a percentage of overall workload, but it's still there. And you mentioned flying cars, so as an example the GPU that's in the Tesla that's doing the real time driving you couldn't send that up to a cloud, no matter how close to the edge it was. I would not want my self- driving car to be reliant on a good LTE connection to-

Avi Freedman: The telemetry is delayed by two seconds it could be okay. When we get to real metaverse ready player one, then we have to solve some application architecture issues, but then you might need sub millisecond locality, but...

Grant Kirkwood: That's right. So that has to be in the car at that point. So if you think about the overall, getting back to what we were talking about before, where there was the universe of IT, which are servers and storage and stuff like this, and then telecommunications to connect it all together. The compute side of it, the consumption model has transformed to this very, sign up online, credit card, launch a virtual machine and go do something with it. And then when I'm done with it, I turn it down and I stop paying for it. Even though underneath all that are very physical assets, so it's buildings and generators and servers, heavy metal stuff. Even though it's very much a capital investment to do all that stuff, the industry has shown us very clearly that you can say, I just want to consume a little bit of that just for a little bit of time, turn it up, turn it down and build a very successful business, doing that. The last mile access business, the traditional telco fiber into tall shiny building has not evolved in the same way. And so I think there's a real opportunity for last mile access providers to shift the way they think about providing access into buildings and into something similar where you're consuming the network services and capacity that you need on demand for the period of time that you need it and not do the traditional, sign the piece of paper for the three year contract at the fixed price for the fixed product that you have for that whole term, whether you need it or not. I think there's a real opportunity for that part of the business to catch up to the way that we've become used to consuming IT compute storage capacity.

Avi Freedman: They still have all the CapEx costs though. So they still have inaudible to keep that building connected unless they have a more flexible back haul. I remember all the people in the nineties that were talking about doing wireless and all that actually just bought T1 tails and lost a lot of money pretending they were doing wireless, but weren't actually. And it's interesting, I talked with Elliot Naas about smaller ISPs and why he's trying to be one, but we have a lot of both direct and we have some wisp aggregators that larger wisp that does what we used to do in networking coopetition. They sell to people that flexibly provide to their neighbors. There are people playing with this, but usually it's more the people that were... If they're not old enough to be hams, they have that mentality, which is communication is good and we need to keep this going. Like what you would imagine could happen in clandestine if the barrier came down to communications for our country, what would people do? They'd probably try to CDM what the digital version of that is, connect things. Is that what you mean? Is it wireless, or is it just a business model? And are you talking about mostly multi- tenant buildings, residents and office?

Grant Kirkwood: Yeah. Yes. Mostly multi- tenant and yeah, it's not a technology problem. It's a business model problem. And, yes, there's still that capital investment that you make to core into a building and put fiber into it. But you could argue that it's the same as Amazon making a capital investment to put servers in their data center and make them consumable in that model. I think it's harder for a telco to do it because it is so locationally defined. Who is in this building? And are those customers sufficient for me to recoup the capital investment? But I think the thing that cloud in general has shown us is that if you make needed resources, easy to consume.

Avi Freedman: crosstalk consumption. Convenience drives consumption.

Grant Kirkwood: That's right. You make it easy to consume, it gets consumed. And telco capacity into tall, shady buildings is still very hard to consume. It's still calling salespeople and negotiating and signing contracts and inaudible. It's still not a crosstalk

Avi Freedman: So it's network as a service for the eyeball edge, whereas network as a service has been... Because I'm very familiar with all the virtual interconnect companies. I did hack at fabric version one when it was just metro area where we're just like, damn, why is it so hard to get between buildings in New York and San Francisco? All these big peering bill buildings are all over the place and you have to dial through the phone if you're in Singapore and order a cross connect and do all this stuff. And then the modern generation is... The non- marketing speak is, they're a telco that automates so you don't hate dealing with them. And that's ultimately what people are trying to do, is telcos and every business starts thinking about customer success and the lessons from SaaS business. So you're saying just do that to the edge. It'd probably be easier if there's more consistency. So US is difficult because as of the regulatory inconsistency. Probably be easier if there's more consistency about access to infrastructure and stuff like that. But that's interesting. So we'll throw the challenge out. Anyone's interested in doing that, reach out to me. I do know some investors that are happy to invest in deep infrastructure, even if you're not tokenizing it and doing packet coin and whatever. Do you have any grand crypto plans for Unitas? You're not going to do packet coin or COLO coin or cloud coin or anything like... Okay.

Grant Kirkwood: Unicoin.

Avi Freedman: Unicoin. Yes. That would be a good one. So how many people is Unitas now?

Grant Kirkwood: It's about 130.

Avi Freedman: Okay. As you look at the industry, one of the things you said to me, which I've talked with other guests about is that the people that wind up building this stuff are people that have been at it for, I'll just say a few decades. How do you think about career progression and getting people into the networking business at Unitas?

Grant Kirkwood: Well, I've observed over the past period of time, 10 years let's call it, that there's not a lot of fresh blood, I hate to use that term, coming into the industry focused on the underlying plumbing of the global network. And if you look around in the communities that we both have spent most of our lives in, it's mostly the same people, we've all changed business cards a bunch of times, but it's really the same group of people that have been doing this for a long time and the need for that plumbing to keep operating and growing and being operated reliably on a global basis is only getting more and more important. The digital economy that we were talking about 20 years ago has very much arrived and the internet is how it works. Everything from going to the coffee shop and using the little striped card to pay for your coffee, with no internet? There is no coffee. Which would be very unfortunate. So I think that the need for making all of that work is as important as it has ever been and continuing to get more important, but it requires that there's the next generation of people coming into it. When I got into this space, it was racks of modems with fans blowing on them to keep them cool. And today it's a very different world. We don't just... Fun story. When I had our first pop in the one Wilshire building running head... There we go. Running-

Avi Freedman: This is my ISP circa 1994 at the UPS next to the water heater, the modems on the back, the CEO cabling run.

Grant Kirkwood: It looks very familiar and running cross connects in one Wilshire there was no service order. There was no portal. It was, you grabbed a spool of cat six and a ladder and probably a drill to go pull that cable. So that part has very much changed, but the need for what it enables has not. And so we need more people to get interested in that piece of it. From a career path standpoint, what I like to encourage is people to get into the network operation side. So like other network providers we have a knock where we have teams of people that monitor the health of the network and monitor customer sites and respond to issues that happen, whether that's a backhoe or something else. And from there, I think there's a natural progression to the engineering and design and architecture of how networks work and interconnect with each other. But, like I said, I think it's important for all of us in the industry to find people that have the intellectual curiosity to understand how does that packet plumbing actually work behind the scenes and get interested in it. And I will say, be the first to say that it's hard to compete with the newer, sexier, the latest, greatest app that you can go and develop and make a bunch of money at it. It's a lot less glamorous, but it's equally, or if not more important.

Avi Freedman: Do you have a way to quantify? We used to call it the bright, shiny eyes versus the dull potato eyes. The people that... It's difficult because this is really something you need to reach back into the educational system in college, especially, and sort of tell people, this is your destiny to own, don't just show up and wait for someone to tell you what to do. But do you try to quantify, the ask questions, explore, figure out... As you see whether it's support or system stuff, or our internal applications or networking these are the tracks. Do you quantify that to people when they're hired or look for curiosity or things like that when you're hiring people?

Grant Kirkwood: It's a tough one to answer. What I would say is that I look for people that have a curiosity and you can often see that manifest itself in ways that you wouldn't have in a typical interview. People that have intellectual curiosity tend to have other interests or rabbit holes as you might call them.

Avi Freedman: Yes. They go deep. You could find it and all of a sudden you're cooking or car or travel or whatever. And you can learn from them and pick a mini religious fight in a constructive way too.

Grant Kirkwood: So, yeah. It's one thing that I took for granted growing up in this community, was that everybody in the community has such varied and interesting and deep interests in things that have nothing to do with the plumbing of the internet. And they tend to be more than a passing interest, but almost a obsessive nerdiness about understanding everything you possibly can about it. But again, those are totally varied and wide, and can be completely outside the gamut of technology or communications. But the people that typically have some interests like that they're very deep on and passionate about can play outside of this tend to be the ones that have the most curiosity about this particular space and then have a self- driven need to fulfill growth in their own career path. That's a really hard thing to quantify. To your question, how do you qualify, quantify that? I don't know.

Avi Freedman: When you're looking for software developers, you can look at their GitHub if they're in a position to do that, or you can ask them to walk you through something like that. But a lot of people that we would love to have come in, might not know that they can fire off a cloud lab or that there's educational material out there. I know inaudible trying to address this, but they might not just not know enough to know what questions to ask. Yes, if you have someone that's purposely gone down the path, you can say, what does your lab look like? It may not be sitting in your house. It might be fired up on DigitalOcean or Linode, or whatever. But, yeah, we think about that. How do you reach out, look for the people that are interested there. Maybe even do a little selling. Anna at PacketFabric talks about going for math or physics nerds, and talking about the internet as a large distributed system with interesting dynamics and harmonics that you can study and tune, but that'll be more... It could be for graduating senior years maybe, but also more senior people that realize that these things are not all separate, that the math of things and the systems that we depend on are... That's interesting to tie those together. I definitely think we could be doing a better job, especially around peering of teaching and explaining, and peering we didn't talk about this, but one of the interesting complexities is, it's not just about the technology, it's also about the business and politics of that. Ideas are welcome. So I guess if someone wanted to be esteemed favorably applying to Unitas and they were younger in career, focus on passion projects in whatever area. Would that be a good way to stand out?

Grant Kirkwood: Yeah, I think so. I happen to be a physics and space nerd, so I'm always looking for people to geek out and talk about black holes with, but, yeah.

Avi Freedman: So was it a galaxy or a black hole that was on its side. I read some article and I was like, I don't really understand why this is news, that there's a black hole that's rotated with respect to the plane. Was it the plane of the solar system that created it, or did you follow that? I missed what that was, but I didn't really understand.

Grant Kirkwood: I didn't follow that one, but I did just see that they found this bubble that's a million light years across that's only in the radio spectrum. And they think it originated at either a black hole or some kind of super, super Nova or something like Uber Nova. I don't know whatever the next term would be. But no, I'm not familiar with the sideways black hole. That sounds interesting though. Maybe an event horizon that is looking at a piece of paper edge on, right?

Avi Freedman: Yeah. Yeah. I understand conceptually it's like, I have a number of those simplifications. The math and physics departments both tried to get me to be a math or a physics major, because I was running their computers and they made me take some specific classes that were helpful for computer science, but it was clear that I was going to be the Salieri, not the Mozart. I could consume, I could understand anything if you give me enough time, but it's not where my passion was. But I do think that there's a lot of musicians that are in software and internet stuff in particular and I think there's got to be some and residences. I'm not a musician so I don't really see that in my head, but physics for sure. The way I was taught physics is very similar to running network because you're always saying, what should it be doing? How is all the stuff derived from the principles? And there's so many god- damned bugs in networking that whether it's cloud networking or whatever it is, it doesn't work the way it should. So you always are on your feet, you're on your toes. You need to understand what it should it be doing so you can say, is this a bug or not? And it does take an interesting personality to enjoy that kind of pedantry.

Grant Kirkwood: I think part of it too, not to go at the risk of going on a tangent, but part of it is system level thinking, that it's a very specific set of disciplines that you have to be very proficient and naturally aligned with system level, thinking. Physics being one of them, certainly. Things like environmental science, where you have thousands of interconnected things that are very much the same, and networking is inherently one of those those things. And that's to the point about who do we look for. That's probably the other one, is that people that think in terms of interconnected network of things and how they interact with each other.

Avi Freedman: So let's take the, it's not necessarily a con to that point, but just for context, I was VP of engineering at one data center company, although really it was the backbone inaudible. And I was CTO at Service Central in inaudible. My understanding is, again, very conceptual of physical data center design. I understand, and I could talk about UPSs and flywheels or whatever, but if you actually ask me to put the levers in the right places and really understand what's going on, if you ask me the physics of why the light bounces around this way, in my head there's a prism. That WDM is a prism, and there's some active stuff or you could do passive. It's possible to do system level thinking from the micro up to the macro with good approximations that map and I don't want to scare people, you need to understand all these different things. You can have working approximations, which is what doctors do for the human body and the internet and stuff like that.

Grant Kirkwood: Totally agree.

Avi Freedman: Okay. And if there are people interested in getting in, please feel free to ping Grant or myself. Grant, how can people reach you and read about Unitas as well?

Grant Kirkwood: Yeah, I would just say our website unitasglobal. com. My email, phone, it's all right there for anybody that wants to reach out any time.

Avi Freedman: Cool. So I'm Avi Freedman. This is Network AF. You can find me on LinkedIn, Twitter @ avifreedman on avi @ kentic. com. And you can find Network AF by googling it, it has a site and on different podcast that you can download and like. Grant, thanks for on and look forward to continuing to see what you build in the infrastructure space and beyond.

Grant Kirkwood: Thanks Avi. Nice chatting with you. Really appreciate it.


Chief Technology Officer and Co-founder of Unitas Global, Grant Kirkwood, joins Network AF to discuss motivations for starting the company and where they're at currently. Avi and Grant talk about what it is like to be a service provider and a solution provider (MSP) in one, and how it plays into what Avi calls the APIfication of networks and IT strategy. Topics throughout the conversation include:

  • Non-software enterprise companies and whether or not they'll adopt more interest in peering
  • Donut Peering and creating better interconnection
  • Building reliable and resilient networks for edge computing
  • Unitas' growth and how Grant thinks about building networking skills internally
  • Looking for people to join the team who are curious

Today's Host

Guest Thumbnail

Avi Freedman

|Kentik CEO and Co-Founder

Today's Guests

Guest Thumbnail

Grant Kirkwood

|Chief Technology Officer, Co-Founder and Director at Unitas Global