Backbone engineering and interconnection with Nina Bargisen

Avi Freedman: On this episode of Network AF, Nina Bargisen joins us and she talks about her career, what's been really helpful for her in terms of mentoring, community, learning, some of her experiences, being both a service provider and working for content providers, doing backbone engineering and interconnection and her opinion on what's hot and where we can go with the community in and around networking. Hi, everybody and welcome to Network AF. I'd like to welcome my friend and fellow networker, Nina Bargisen. And Nina, could you give a quick intro?

Nina Bargisen: Absolutely. Hi. Yeah, I'm Nina. I've been working in this internet industry for the past two decades or maybe a little bit more. I'm in my fifties now, braving mid- life crisis and have been changing jobs a couple of times over the past couple of years, but started out as a mathematician, going to the university, thought I would be doing research, realized I wasn't smart enough, or at least in that right way, to go that way. And then after having kids, ended up in the internet industry by sheer luck. But what a journey.

Avi Freedman: Cool. So how did you get into networking? How did you find from math into networking?

Nina Bargisen: It was actually, again, as I say, it was a little bit of luck or a big deal of luck. I had kids and stayed home for some years and then when I wanted to get back to work in'99, I landed a job at Tele Danmark, as it was called at that time, in a sort of a intro program for newly graduates. I was five years past my graduation, but since I hadn't worked, I applied anyways and said," Hey, I haven't used my education, so consider me a graduate please and take me into the program." And they did. And I think here is the first element of luck, deciding to do that in November where I applied in July 1999. Had I waited another year-

Avi Freedman: The bubble, yeah.

Nina Bargisen: Boom. The bubble would've gone and they would probably not have hired 30 people into these training positions that they did. But lucky for me, I just made it and then quickly moved... We were supposed to be different areas of the company. I started out in telephony, in a strategy position, thinking about IM telephony, so the telephony services that you could punch on your pads and you would get a service. Call back or call forwarding. Some crosstalk-

Avi Freedman: So even the telcos were thinking about-

Nina Bargisen: crosstalk.

Avi Freedman: ...Services that added value.

Nina Bargisen: Value- added services, yeah. And they were run by a big computer. And so I was ended up in a project trying to figure out whether it would be a good idea to combine the platform that was used for the mobile network and for the fixed line network, because of course that was run on two different computers because it was very separate networks at that time. Probably still is. So I wasn't there for very long and then I moved on to the internet division of the company and I started working with streaming products in 2000. So that would be-

Avi Freedman: Oh, that's pretty early. Yeah.

Nina Bargisen: That's pretty early. Yeah. We were trying to build like a enterprise products, so people could host video advertising on or just also just having shared hosting of video files and ended up being involved in the first runs of the Big Brother crosstalk-

Avi Freedman: Wow. That was back then? I guess I'm old too. I'm in my fifties. Yeah.

Nina Bargisen: I know. That was... What a horrible concept, right? And it was a big thing and it definitely... a lot of us learned a lot about live streaming, a lot about capacity. In particular, because we had that set up running, I think, for the second time around in 2001 when September 11 happened. So I remember when that happened and we were watching CNN on our computers in the office. I got a call from one of the TV stations that we had been running their test or their first initial live streaming of crosstalk that they were doing. And my contact, he called me and he said," We want live streaming right now because this shit is going on." And very quickly we got everything directed around to the Big Brother set up. So sort of like," Sorry, Big Brother, we kind of need these servers for something else." So we did one of the local TV stations live streaming. And I remember getting a picture a couple of weeks later, somebody in Australia had been watching the live feed from this Danish TV station on the internet in Australia. A Dane who lived there. So that was one of the first big streaming events that I took part of.

Avi Freedman: Yeah. I was at Akamai at the time and we were fortunate that we'd had some big events, Steve Jobs, Victoria's Secret, but ironically Danny Lewin who was Akamai CTO and one of the co- founders was really trying to move Akamai from just objects to whole site delivery. And he was the first person murdered by the terrorists on 9/ 11 on the first plane. But all the people that were disagreeing with him about that needed to call Akamai and say like," No. We need to do a whole site right now. Our load bouncers, our SSL, the bandwidth, everything, our databases, everything is falling over." So it was a proof, but yeah, those were crazy times and obviously everyone at Akamai remembers what was going on then. What was after the streaming group at TDC?

Nina Bargisen: After the streaming group I had a short stay in the operations and then I moved on to the internet backbone planning group. So what happened was that the IP was growing up. The IP group had been one single group in the research and development group and at some point, at that point in the early 2000s it was decided IP will now be important enough that it should be structured like the rest of the company with a separate engineering, a separate operations and a separate planning group. So they split up that group and hired some more people in and I ended up in the planning group.

Avi Freedman: Was that internal, external, both?

Nina Bargisen: It was internally, but this was then in the fixed line company. At that time Telemark Danmark was called TDC and split up into an internet company, a mobile company and a fixed line company crosstalk-

Avi Freedman: Was the group more focused on backbone interconnection or both? Was it joined?

Nina Bargisen: So I started out just focusing on edge, like B- RAS planning, but then after I did a good job and they were like," Oh, Nina, when people are doing a good job, instead of giving them money, you can go NANOG, this North American thing." And I went to my first NANOG in Reston and it must have been 2002 or 2003 or something, I can't even remember, and there I met the peering crowd because there was a Terramark peering forum. So in a separate hotel and I remember they," Nina come to this meeting." I met Ren Provo. I met Patrick, I met some of the people and they was like," Nina, you should come to that." And I even, at my first," Okay, I'll go." I went up the stairs and I said," Oh yeah, I have this AS number. We have a pop in New York here with us." I had no idea what I was doing. So when I came home, I went in to my manager and I said," I want to be the peering coordinator instead of the peering coordinator." And again, luck was there because Ken, my manager was," Hey, awesome." Because Michael, who at that time was the peering coordinator really wanted to go back to the engineering group. So he was one of the engineers and he had been very focused... He's a very sharp engineer but he had also been very focused on the interconnections and on backbone designs. So when they did the split, they put him in the capacity planning group, which he was not happy about at all. So he really, really wanted to go back. And then there was this problem, " What are we going to do about the peering part? Because Michael is the best for that." And then I came back and I said, " I want to be peering coordinator." And he said, " Awesome. Nina, you can do the peering coordinator."

Avi Freedman: At the time, what we would now call selective or restrictive peering in Denmark, but open peering in New York, was that the strategy at the time?

Nina Bargisen: Yeah, that was very much... the policy that was developed at that point was restrictive to the point of always saying no locally, because otherwise a wholesale business that they," We want to sell transit, so we can't really peer with all these small networks in Denmark, because then they won't pay us money for the internet."

Avi Freedman: I was at a peering thing in Asia Pacific, back when I was at AboveNet and we had bought... I think it was we bought PAIX and one of the incumbents was saying," Oh, it's so horrible. We come all the way to the US and AT& T won't peer with us. They say,'you look like a customer.'" And I said," Oh, Jean, how many people, how many networks do you pair with in your country?" He said," Well, that's different. They're our customers." It's like," Oh, okay. I see." You're my customer. So I won't say anything about that, but it's funny because at Akamai I had to teach people," You need to let people be wrong and make it make sense in their frame of reference." Some of these big companies have their own frame of reference. So like if someone says," We do not pay for bandwidth." It's like," Well, maybe we'll give you space instead with some bandwidth." It's like you see that, right? And by the way, Nina. So I, when I was trying to figure out this BGP stuff, in 1995 or so, it was a big," Shh, don't talk about it." And I just sent mail to of the MAE- East admins. I'm like," How does this peering thing work? Do you just like when you get a based IP address does peering happen?" I didn't even know about multi- lateral peering and agreements and route servers and stuff like that. There's a lot more data now, but it can be very frustrating. So, thank you for helping grow the community and take new people. So, when you were getting started, it sounded like you got lucky that you came in during a time of big growth and it sounds like TDC had a formal program for bringing people in and mentoring them. What was really helpful inside the company early in career to getting access, getting exposure? Was it people, processes, and what, maybe, could have been better that you think about as you hire people and bring them in now?

Nina Bargisen: I think one of the really great things about the IP group at TDC and that was very much in the fixed line, the backbone group, I had actually met the exact opposite in the internet group. It was very closed, protective." I don't want to share my knowledge because then I can keep my value up." Where in the backroom group, it was way more open. So when I showed up not really with any formal engineering skills and go,"That BGP stuff, how does that even work?" I want to be the peering coordinator instead of the peering coordinator, but I kind of need to learn this BGP thing. And I did take a class like a Cisco BGP introduction class, but learning it, really learning it, really understanding it, it was a lot of crosstalk-

Avi Freedman: The dynamics of BGP are interesting.

Nina Bargisen: Exactly. And what happens and how do you do traffic engineering? How can you move your traffic around inbound, outbound? And all of that took me a little while to grok, but here I met some amazing engineers and Michael, the former peering coordinator, is actually is one of them who never was too busy to teach and to sort of go," And, okay, here's your access? Now you have a login on the routers. Here's the testnet. Maybe you go do things in the testnet before you try them out in production and let us know." So they were very open at that time for people who were interested in learning to learn and grow also out of their more formal responsibilities, just because," Hey, somebody likes to do this shit. Let us teach them."

Avi Freedman: Of course there were more serious router bugs back in the late'90s and then the early'00s. Was there more freedom and permission to fail, experiment? Or was there already rigorous like," We have a testnet. We test deployments, we test code."? As a telco, did they have sort of that mentality that you could experiment, but test rigorously before deploying?

Nina Bargisen: I wasn't really involved in that much in testing code, but they usually did that in the testnet, but we also had this internal joke, there was a small testnet and the big testnet and that the things you would do, you wouldn't really find all the bugs before you had it out in the big testnet.

Avi Freedman: Well, it's hard to test traffic engineering on a prefix that's on a peer. It sounds like a Dr. Seuss. A packet on a pocket on a prefix on a peer.

Nina Bargisen: Exactly. And I think back then there was a higher tolerance. It took a while and probably also a number of big breakdowns before a change management process was introduced, before service windows. For a long time, I was working with people where it's like," Well, we should do the changes while we are at work, instead of doing them when we have to wake somebody off to get shit fixed." But that definitely was something that changed over time. And in particular, when more and more of the... When I started out there, early'00s, it wasn't really very important services running on the internet. Once TV started running on the internet, once telephony started running on the same backbone, once we had customers like CSC, I think were called at the time, that was running big enterprises IT systems. I remember I did some network changes that was close to being non... It shouldn't be bothering anybody, but I kept all SAS's planes on the ground for an hour.

Avi Freedman: Oops. Yeah. It's a great debate. In my ISP days, they used to have a bat light that would go off when I enabled. And my point at the time was," Well, if I'm making the change, then we might as well..." Yeah, as you said," Do it while I'm awake." But there was a point where I stopped pulling the plug on UPSes with the same theory, when I happened to be at a random POP and customers gave feedback that that was not the way they wanted to work. And then I enjoyed being at Akamai where, if there was a problem, it was a software problem, not a networking problem, so that was awesome. The network, it was food, but not the problem. So, you talked about the peering community and your entree to that. What was it that... because interconnection and peering, it's certainly technical, right? You have to think about traffic and peering and it's also people, what are their motivations? And business and analytics, you have to make a case, but when you actually get to implement things, it's," Woosh." The cannons move and the traffic moves. There's positives and negatives. What was the compare and contrast and what was it that really made you more interested on the interconnection and people side?

Nina Bargisen: It was a combination, I think. It was the community. I fell in love with the community. It definitely has its weirdness. I like quirky, weird people. I liked the idea. I got introduced to the IRC channels where people were hanging and that whole community and that whole bubbly communication and helping each other, I really fell in love with that. And then also just handling the traffic, man. Deciding," We are going to have that traffic in there and that traffic over there." I thought that was amazing. And in combination with being in a planning group and sort of," How can we build the network most efficiently?" Also meaning how can we get the traffic into the network most efficiently and cheapest. Do we want to buy everything in Copenhagen or, at that time, we were running on a network that somebody else had built out so into Europe and we had those connections into the US and we managed to be able to build the case every time we needed an upgrade, is it a better case to spend the money on the upgrade, rent some more capacity, stay in the pubs or-

Avi Freedman: Because people externally-

Nina Bargisen: ...Dismantleeverything and just buy some transit in Copenhagen. And every time the case actually came out, it's a better deal to keep the network up and we would have to pay for that. So we might as well build our own network.

Avi Freedman: So there was people engineering, internally and externally.

Nina Bargisen: Externally. Yes. There were business cases to be made internally. And then the negotiations externally and then just the whole," How do we get the traffic in?" Like another big streaming event was when that guy jumped into space.

Avi Freedman: Parachuted from space?

Nina Bargisen: Yeah, it was-

Avi Freedman: It was like-

Nina Bargisen: What was this?

Avi Freedman: Six years ago?

Nina Bargisen: In 2007, I think. And it was Red Bull, they sponsored this guy who flew up in a balloon almost up to the edge of the atmosphere and then he jumped out and it was live streamed. And this was in 2011, I remember. It was the year before Netflix launched because I remember it was so much extra traffic and it was coming in on all of our... We had Google caches, we had inaudible caches. It came everywhere. It came in on every transit connection we have, on every peering connection we have, like everywhere there was a source. And I remember me and I remember my colleague in Finland, he was frantic because he saw that all the traffic to Finland was coming out of New York and he was like," You got to do something. This is not fair." And I pinged my friends at YouTube and was sort of like," Why is it coming from New York?" And he is like," There's nothing we can do. No more servers near Finland with any capacity, so it's either that or they're not getting it." So I thought that was another big event. And then the year after Netflix launched and that amount of traffic that we was sort of like," Oh my God. We have so much extra traffic." That was just another day when Netflix had opened up.

Avi Freedman: I guess we shouldn't speak ill of the former networks, but I was at Akamai, and I forget whether it was Steve Jobs or Victoria's Secret, but I had a call it at home because it was the same thing. It was like we filled every port of at home, up their back, in the front, in the side, everything. And we said like," Don't you want to have more capacity, take some servers?" Like," No, no, no, this is our strategy is to congest so people need to pay to have servers inside." And I was like," Well, we're not going to cooperate with that strategy." So they're on my list. Aegis was a network that while they were a spam haven, so we said, a global internet, spam haven, but they were one of the first networks that at NANOG, I think it was an Ann Arbor one, got up and described their peering policy, which they themselves did not qualify for, which was a funny thing. So hopefully the good enlightened networks will survive and I would say the internet's and its content and everything has gotten a lot better, so that's pretty cool. Speaking of content, you have worked for a service provider and then as Netflix started back then, but you went to work for them. What was it like? Interesting, more interesting, less interesting, different? And role- wise, the differences of doing interconnection for... at Akamai called it" the world's largest non- network" because there's no backbone and stuff like that. So it was actually Rob inaudible, who... No, no, it was Andrew Koo who called it that. It was Andrew Koo who called it that, but what was it like, what was the difference on the content versus access and transit side?

Nina Bargisen: The big difference was definitely now we would peer with everyone except for very, very few people whose behavior meant that it would mess up our traffic more than it would help to peer. And because working for a service provider and doing peering there's this whole balance about," Could we sell to these people?" Or if we have a division who wants to sell transit, so we couldn't have a completely open peering policy because... or maybe we could. The idea was that we couldn't. So we had to protect the wholesale business. And there was a lot of, we could do trade offs like," I buy some waves, I get some peering." Trying to do that whole deal. TDC is probably a little bit too small to really do anything interesting in that field, but we were still trying to do it.

Avi Freedman: People bought data centers and data centers cleaning services from AT& T to get peering.

Nina Bargisen: Yeah. And also we had business customers, a few very important business customers, so we could not do too many little tricks with traffic or hairpin it too much, because then we would have customers who would be upset. And there was some customers that we really cared about and then there was, well," It's just internet, if it's a little bit delayed, doesn't really matter." And we could play with, with that kind of traffic to get the peerings that we wanted. It would work sometimes and sometimes it didn't.

Avi Freedman: Yep. No, that's a grand question is," Who cares about their customers more?" And that.

Nina Bargisen: Exactly. And with going for content, obviously we cared a lot about getting close, but then again, on the other hand, video is not really latency sensitive. So that traffic could be hauled around a little bit if that was better for us. But in general, we were open and we were on the other side, it's like," Hey, eyeballs." We really want to connect. But very quickly, also, in Europe, Netflix became important enough that the eyeball companies was also the ones with," Hey, Netflix, can we somehow connect?" And with the embed program that... I started out just doing interconnection, but quickly ended up doing embed program as well. There was a lot of handling small... trying to explain why we couldn't really send them a server because their 10 megs or their 20 megs of the Netflix traffic would then be replaced by 400 megs of inaudible traffic. So not really a great idea. So the role changed very much that way. And also not having a backbone was really weird in the beginning. It was like," They have a backbone now, the Netflix."

Avi Freedman: Right. The world's largest non- network, right. Akamai has a backbone now, too. Yeah.

Nina Bargisen: So, but for many years we did not build a network, a backbone and when we did, it wasn't called... it was the inaudible network and not a backbone. I think they call it-

Avi Freedman: inaudible rsync. Yeah. It was to rsync everything. Yeah,

Nina Bargisen: Yeah. No, actually, Netflix doesn't do that. Its servers are working as clients and they just get a," You need to have these movies on." And then they request the movies.

Avi Freedman: Okay, so it's not push, it's pull. Yeah. I know Netflix caused problems for Akamai when I was... And maybe it was after I left because Netflix can do overnight sync, whereas Akamai has to be live. A lot of the main CDNs need to be live because, as you said, one of the really brilliant things I thought about Netflix, which I don't know whether it was brilliance by design or sort of accidental, was by having the intelligence in the client and having adaptive bitrate and coding, you were already going to be hunting and saying," Can I get 10 megabit, four megabit, one megabit, whatever?" And then you could go from local to worst peering, cheapest to most expensive transit before failing down to the next and do that all desynchronized. So it was so much easier than being a centralized CDN, and then yeah, the push or," Please go get this popular content." So much easier than caching. Not that... I know the optimization of servers doing hundred gig plus was huge, but some of the life was easier and made life difficult for CDNs. So like,"Can't you be like, Netflix? Can't you be like Netflix?"

Nina Bargisen: I know. And you know, they will love hearing this, I'm sure. But again, I think it's by design, having the intelligence in the client. Having a client is key to a lot of things that Netflix has been successful with, because when you have the client, you have the end to end control and you have the end to end data and, well, where we are right now, data is key. They've done well. It was a great journey.

Avi Freedman: Yeah, no, I've been impressed. Actually, I saw Netflix first... I forget where I was, but I met Vish when he was there, when they had built out a whole set top box plan and data centers. And then, again, it was an interesting look before the whole culture thing came out and Netflix talked about how they want to be, because it was like Reed woke up and said," That's not what we're doing." And just like," Nope, that's not the right way. We're going to do streaming." And I guess it was right. Yeah. And the content stuff too, so definitely some interesting business lessons there. But yeah, when I was at Akamai, I was interesting because not every network person comes from a background of being adept at making those sort of business arguments. And it's like as an inverse sales, where you're trying to get people to take things and then, as you said, maybe let people down gently. Although I definitely wasted Akamai money, not only trying to build satellite network, because we didn't know how big and how edgy it was going to be. We were off by about 15 years, maybe 25 years about the true edge, because we're not at the true edge yet. And yeah, we actually set 10 megabits because I was like," It's all going to grow. It's all going to grow." crosstalk thousand and everything was going like that." So we were like," Sure, 10 megabits sure. We'll send you$ 20,000 of servers and switches." They didn't fire me for it, but who knew, but there was a point at which I was signing... Like it took at least half a day a week to sign all the orders for everything. And oh my God, taxes. Did they have all that figured out at Netflix by the time you joined? All the in country taxes and stuff like that?

Nina Bargisen: So luckily I was EMEA, right?

Avi Freedman: Okay. Yeah, crosstalk is pretty bad tax wise.

Nina Bargisen: Yeah. Europe was good. The EU was easy to ship to, but once you started shipping into Africa and taxes and all of that, but they worked it out. Learning from being active in Brazil pretty early, things had to be very, very bad to beat Brazil when it came to taxes and shipping and stuff. So it was all like," Ah, as long as it's not Brazil, we'll figure it out."

Avi Freedman: We also, there was some mandates, I think, that you needed to buy servers that were built or assembled and then Dell had a plant there and at Kentik we actually have a lot of Brazilian ISPs. Someone's making routers and making them available there, because there's a lot of interconnection down there. I would love to go back-

Nina Bargisen: Yeah, no, you can import into Brazil now. We managed to import, but we had to have a Brazilian entity who did the import. Most of the time Netflix donates the server to the partner. In all of the countries where that makes sense and is allowed tax wise and stuff, that's the best model to use.

Avi Freedman: They thought I was kidding when I was talking about laptop regions at the time, but this was before SSDs, so we didn't really have the IOPS to like go were down with suitcases. And we weren't going to do anything, we were a public company already, so we weren't going to do anything that was not regulatorily okay. But yeah, I remember we had people installing, you mentioned it, IRC servers or Usenet on some of our Akamai servers before we had some of the auditing when we started doing SSL. And then quickly, I think, within a few months of when I was at Akamai, we had a lockdown, so if someone did root it and did something, it would disconnect from the network and we'd contact them. And we even had a couple cases of people stealing the servers and posting them on eBay, thinking that Akamai made servers. So they'd show the picture with the serial number of a stolen server on eBay, which was always interesting.

Nina Bargisen: Oh, that's brilliant.

Avi Freedman: So, did you-

Nina Bargisen: I'm not sure. I don't think... At my time in Netflix, I don't think we saw any Netflix servers on eBay or anything that was stolen. And we lost very few. I think we lost some, but not anything that we were thinking about. It's a funny story. We did have someone call us from a data center once and say," Oh my God. Somebody messed with your servers. We're so sorry. We don't know how it happened." And that was because, at that time, we would paint movie quotes on the servers and this was then a cluster that was built, one of our own clusters in a data center. And inaudible for some reason, the person who were doing the quotes were in a funny mood and were doing all kind of sexy, little bit naughty quotes. So we had these two racks of servers with naughty quotes written in silver on it. And then when this data center technician saw that, was like," Oh my God. Somebody wrote naughty stuff on Netflix servers. What are we going to do?"

Avi Freedman: Oh, they thought-

Nina Bargisen: They called us and said," We're so sorry. We don't know how it happened." And we're like," No, no, no. We did that."

Avi Freedman: At AboveNet, we had Cisco... Not ASRs. What were they called? Backbone routers. And you could make the line cards, you could program what they said. So actually, for a while, we programmed them to flash our AS, but... I remember the acronym, speaking of naughty, people called them BFRs at the time, but that was the code name for it. But I forget what the product name was, not CSR. Anyway. So for a while, then someone asked permission and flashed ABOV, AboveNet, our stock ticker and the stock price, which worked okay until the bubble and then they took that off. But that was like the confluence of... And of course now we're in a bubble again. So, yeah, it's interesting, definitely different. Same ideas, whether you're engineering the packets to work internally or doing the negotiations and setting policy and planning, whether you're access, hosting, transit or content, making things go. And to the average person, it's all the magic of the internet behind the scenes. Personally, I am generally fascinated by all of it and definitely the people side can be a little bit of a challenge. And again, how do you help someone see that what you need is really what they need too, but their terminology and way of thinking about it is not there. And some technologists get very pedantic, like geek binary- itis like," No, no, no, but they're wrong. They must be shown that they're wrong." It's like," No, no, that's not only the way to influence people." So, yeah. Cool. Well, thank you for sharing. So you've been, as you said, following networking for some time. What's changing, what's hype and what's hot? What's the things that people are talking about that aren't here yet? And what's the things that you're excited about in networking overall right now?

Nina Bargisen: I think it's the new edge, like things being pushed closer and closer. For years, we've been saying it's all slideware. 5G is probably just going to be another 4G, but it looks like people are actually more serious about putting compute closer to the edge. I think a lot of people think that a lot of the content will come to the edge as well. At least I doubt that video- on- demand is going to be on the edge. Live's probably... I can see live would use the edge data centers and servers sitting on the edge, but on demand, it's not enough files, it's not storage place. For the video- on- demand server to be efficient, you have to have a high demand. So there's no use putting that next to the mast, because that mast is never going to serve enough users for that server to be efficient.

Avi Freedman: And there's been a wave of problems with power density of putting processors in routers and-

Nina Bargisen: There's so much. So I think the non- latency sensitive services, like on demand video and a lot of other... Like pictures and everything's going to stay relatively central, not totally central because it is decentralized, but it really takes a lot of users to be efficient and I think a lot of the real time, edge compute kind of stuff, you need to be even closer than where the video files are. So I'm-

Avi Freedman: You're optimistic for the edge and all the crosstalk.

Nina Bargisen: I'm optimistic. I'm cautiously optimistic because I've seen some happen, but also seeing some of the edge hosting companies ending up," Yeah, we want to do edge hosting." And" Yeah, we want to do something else."" Yeah, we don't want to do anything at all." So it still remains a little bit to be seen, but now that the actual networks are coming up and the slicing is starting to happen, maybe there will be some changes in that respect because in the past eight years where I've been mostly focused on broadband internet, cloud has really changed the internet a lot. So I'm actually on a little bit of a learning journey right now to fully understand how the enterprises are running their stuff and how the software services running their stuff in the cloud. Because that happened while I was looking on the other way, I was more focused on," Oh, all these program providers. Here's some video traffic. What are we going to do about that?"

Avi Freedman: Also interconnection, right? You can now, for the enterprise, you could interconnect through your SD- WAN. You can do an old school, you can get Equinix cabinet and run a cross connect and do peering and do an IX.

Nina Bargisen: Or you can crosstalk.

Avi Freedman: You can use PacketFabric, Megaport, PCW. You could, I don't know, soon there'll be ESP or something. There's lots of different ways of-

Nina Bargisen: So many different ways, yeah.

Avi Freedman: It's like Star Trek, IDIC: infinite diversity in infinite combinations. crosstalk

Nina Bargisen: And then again, I can't really figure out, is there really some magic or is it just private networking and some software? And I think probably the last, but it's presented as a lot of magic. So I'm still trying to-

Avi Freedman: I'm a big fan of the innovation and companies in the interconnections space. So again, PacketFabric, Megaport, PCW, but I do have to laugh when people talk about a wave of companies as network as a service. So what was TDC? What was AboveNet? What was... We weren't doing packets as a service before? But, you got to have marketing and easy ways of people to remember things. inaudible has sassy, we have network as... and NAS, network as a service. Yeah, the edge, it's definitely a lot of innovation there. I'm really curious to see, also because the dirty little secret is that the edge is some pretty big megawatt data centers where, where people are actually served from, whether it's cloud. I like going up to-

Nina Bargisen: That's where it is right now, but does it have to be more distributed than that?

Avi Freedman: No. Yeah.

Nina Bargisen: How much does the milliseconds actually count?

Avi Freedman: And for what? Ready Player One? When we're Ready Player One, milliseconds will probably matter. We're not quite there yet, but yeah.

Nina Bargisen: Yeah. And then know with the... I don't know, I'm still waiting for the self- driving cars. That runs on some network kind of thing and make sure that they're not running into each other, or I don't know how they-

Avi Freedman: I like networks.

Nina Bargisen: I'm not sure I'm comfortable with that.

Avi Freedman: I'm not sure I would be in a car which might go off a cliff if it lost connectivity to Google, but there's certainly a lot of telemetry. There's certainly a lot of real and IoT with doing model development things at the edge and using power that you're using anyway. So yeah, it'll definitely be interesting to see how that develops. Maybe we'll do this again in a few years and see what the world is like. You talked about getting into the people side, both inside TDC and when you're at your first NANOG and then probably RIPE and APRICOT and Global Peering Forum and European Peering Forum. And it sounds like you had pretty welcoming and pretty good intro. I'm curious, over the course of your career, did you find that to be generally true? Did you find that all the communities were pretty welcome, pretty educational, or are there areas that you hit frustrations in that you have constructive criticism for us as a community about ways we can do better?

Nina Bargisen: I think the peering crowd was a pretty welcoming crowd and I think it kind of comes with the job, right? You have to be able to talk to everyone and if you meet somebody who's interested and a potential, it seems like to be welcoming... I think the engineering crowd, that's the non peering people at NANOG, was probably not as welcoming and more difficult to get to talk to. And there definitely have been some... I remember a dinner where I went home and I was kind of crushed because I was sitting next to a person and he was very interested in one of the hotshots who was sitting next to him on the other side. And so I was at the end of the table and I was just sitting there and just had this back turned me during the whole dinner. So like a 10 people table. And I was just like," Whoa. I do not like that." That was... so I think it can be a pretty tough crowd sometimes if you don't know somebody. Usually I went to all these events by myself, but over the years, you find the people that you get along with, you meet friends. And again, I learned so many things just chatting with people who were awake at the same time and I would be up late at night and talking to people that would teach me stuff.

Avi Freedman: Oh, that's right. Your morning would be the normal engineering hours of people still up.

Nina Bargisen: Exactly. So I could get up in the morning and I could talk to somebody, a drunk RAS and persuade him to explain this MPLS to me, because I was in a big debate with one of my engineers in the engineering group about something and needed," I don't think it's like this, RAS. Can you explain this to me?" And then I would have a little bit more ammo in the discussions because I learned it from... I didn't learn everything from the TDC engineers, I learned from outside as well, so that made it even better to get more inputs into... When you're basically... because all of I learned everything just by mentoring and working it out myself.

Avi Freedman: The IRC has never fit well for my brain, but I've talked to a lot of people that... I know a lot of people that are tight with that community and it's seems like it's gone really well there. So it has not moved to Slack. It's still IRC on the coordinator.

Nina Bargisen: There might be some Slack. Some people might have moved to Slack and I didn't notice. I've been working most of the time over the past years but when I started new job last year I needed to get some old contacts and I couldn't remember their email addresses and sort of like,"# IX" and yeah, these old guys are still hanging there. Sort of like," Can we talk about this?"

Avi Freedman: Do you have IRC on your phone? Do you have a phone IRC client? Or is it really laptop crosstalk?

Nina Bargisen: No, it's on the laptop, it's..

Avi Freedman: So not all the way. I guess Discord would give advantages there. Would you say it's more than 50%, more than 75%, of what you've learned has been mentoring, osmosis, independent experimentation? Like how would you rate the sort of the tribal explanations versus the formal and is that something we could do better at formalizing, especially in a COVID world or post- COVID?

Nina Bargisen: Well, for my personal experience, it's like 80%, 90% tribal and self- study and working it out. But a lot of the working it out is also then hitting up somebody and said," I thought about this, blah, blah, blah, blah, blah." And then you have... So it's definitely tribal. And because I have that experience, I think it's difficult for me to point out whether we could formalize that and make it more available for everybody. We kind of should. And I know there are some programs that are teaching networking at universities, but you often end up... I know I actually had a nephew who wanted to become a network engineer and he started this class and it was a joint program with app development. And within the first six months of this 18 month program, everything was turned into app because they had counted on Cisco sponsoring and making them CCNAs or maybe the next level and apparently that kind of fell through. And then the teachers were like,"Oh, well we don't know anything about networking, but making apps is more fun." So my nephew ended up dropping out because he was like,"But I kind of like this network thing." And wasn't encouraged to go find an apprentice place or anything because he didn't get the support from the school. So I thought that was kind of sad. That was a local thing here in Denmark, but I don't know that there is a network engineer education here anymore.

Avi Freedman: Yeah. And in the US, at least a lot of network education is still distance vector protocol, prefix, NLRI, what offsets things are. It's TCP books and pulling packets apart and not the dynamics or it's an economics group that is trying to understand how does the internet work and how do people connect with each other or... Those people are more likely to go into policy than engineering. It's an interesting question. You mention your nephew. What would you advise people who think that the internet and the infrastructure is interesting to get into that world?

Nina Bargisen: Get the job. Maybe not care so much about getting a formal education first because it can be difficult to find a formal education that actually fits. But make sure that you have basic engineering skills, that you know how to work shit out and then go into a job with the expectations that you have to keep learning all the time because things are changing all the time. To me that's a general thing, when I talk to young people about," What should I go study? Should I go to the university? Yes or no?" I don't know. Yeah, you should go study. You should go figure out how you learn stuff. And you definitely do it with learn something that you care about because the most important skill is to figure out how you figure stuff out.

Avi Freedman: Yeah. I think of it as getting comfortable with getting confused and un- confusing yourself. Ask for help.

Nina Bargisen: Exactly. A very good way of putting it.

Avi Freedman: Ask for help. Be on IRC. Have a community. But yeah, it's something I think we need to get earlier in at least college education about the" find your passion" part of it and that it's interactive. It's not going to be sitting there and getting the received wisdom of the things.

Nina Bargisen: No, no. Exactly. It's actually working with it and figuring out," Oh, I don't know this. I have to figure out how to do this in order to do this longer, far away thing that I have to do." And then go figure out how to do it.

Avi Freedman: Which is why it's cool. There's network simulators. There's virtual appliances. You can get a virtual MX, a virtual ASR, a virtual Arista. And yeah, it's something we want to do, I want to do is try to see what we can do to work with academia, education. I saw something at NANOG a few years ago, which I thought was pretty cool, which was a peering simulator. Not like the DrPeering game, but actual people got to have some prefixes and peer with each other. It was from European region, I forget which. That might be interesting to do with people. Have to go find that link and look at it. So any asks or advice for the community about trying to be welcoming to people that do not look like all the people in the community now. The backbone engineering is more people that look like me. The interconnection seems to be a little bit more diverse, but not matching the population, certainly. What could people stop doing and start doing?

Nina Bargisen: I think the general... Well, having worked for American companies who took diversity and inclusion very, very, very important. I've been trained quite a bit, but I think realizing that the person standing next to you have a very different background, but might know as much as you do, I think that understanding is crucial. And it's very difficult to have. I still do it myself when I meet another female and in particular, if she's wearing a dress and high heels, I am not assuming that we could go and have a conversation about traffic engineering and I hate myself for that.

Avi Freedman: Unconscious bias is hard to escape crosstalk-

Nina Bargisen: Unconscious bias, it's a thing to think about. And I think the young people, at least the young people that I know are much more critical than I was at that age. My eldest child is a queer and just had a top operation done this year. And I remember talking to them about," But how do you think you will feel when you go to the beach and everybody will see that your body is now matching your perception of how you want your body to look, but it's very different from what everybody else is expecting it to look like, because it'll be male on top and female on the bottom and people will look more at you than they do now, where you yourself is in conflict with how you look."

Avi Freedman: And they said,"That's their problem?"

Nina Bargisen: Exactly. The reply I got from them was," Well, exactly. I have been fighting this fight for the past 10 years. Now it's your problem. I don't care, because now I look like I feel." And they're very different that way. So I think in order to include them, we have to make sure that we don't really care about color or gender perception or hair color or what shoes they're wearing or anything. We just have to get past the looks and past the also how you talk and then just get to the substance. But it's a basic human thing, but I think it is important also to be mindful about with all of us, because we haven't been it for so long. It's been a bit of a rough community sometimes, right?

Avi Freedman: Yeah. It has. And sometimes people-

Nina Bargisen: With a lot of drinking.

Avi Freedman: Yeah, drinking. There's been... I can't take cigarette smoke and there's lap dances, there's women that ask people to go to a place crosstalk-

Nina Bargisen: Yeah and ending up-

Avi Freedman: Maybe not anymore. And so you do have to be aware. I think what works best is when people show that they're trying to learn and understand, then there's more grace and when there's less-

Nina Bargisen: We can't be... Nothing is going to be... everybody's not going to be changed and everybody will make mistakes, right? But being open about that we actually trying to learn and being respectful and that we can say sorry when we are not respectful and somebody tells us that we are not respectful.

Avi Freedman: Well, thank you, by the way, for four or five times saying," What about the Europeans at Kentik?" I think we do an okay job of it, but I think there's been a couple times where there's been the... You've been comfortable speaking up that," Hey, what about..." Time zones or people or things like that.

Nina Bargisen: Well, I have to say, I want to say thank you on the other round, because this is the first time where I see regular meetings being started at 7: 00 AM California time. That has not happened before.

Avi Freedman: Oh. Interesting. So it's been more like," Sorry you're having dinner, but we need to meet now?"

Nina Bargisen: Exactly." Oh, sorry, you're just going to miss this meeting." And yeah, you can... it was never really offered that people could get up early instead of everybody just staying up late or other people from other time zones would get up really early to-

Avi Freedman: Well, even there we could do better. I know we had our last all hands a little late. We're working on it. We're talking about it.

Nina Bargisen: Yeah. I just didn't show up. I watched the video. That's fine. As long as you... that's the great thing about that we're running everything on video is that there's a recording and obviously you can't have the interaction, but I could call you up afterwards and say," Oh, well you said this. What do you mean?" And I wouldn't be able to call you out publicly about not being nice to Europeans but I can do that other times.

Avi Freedman: No. Well I think you've been comfortable in more than one person settings doing it. So I appreciate that. But you also have a career in community and all that, so you're in a position where you know that... But we appreciate it, so thank you. Any advice for your younger self? Any things that you would have said to have had maybe either guide different experiences as you were going through the career?

Nina Bargisen: Probably change jobs a little bit earlier than I did. I was at TDC for 13 years and the past couple of years I was there, I wasn't happy. I'd grown out of the role. And a thing, yes, for young people to might be mindful of is that if they land, and I was really lucky, I was landing a role that I could grow, but growing a role usually means that you are not necessarily growing the formal responsibility and the salary to go with it. So what I was really responsible for and the work I was doing was not at all matched with what people a couple of levels away in the organization thought that I was supposed to be doing and who didn't know what I was doing. And definitely my salary was half of what it should have been. And that was because I didn't really care about salary for a lot of years. It was more about I loved work and I just thought that," Well we only have 2%. So we can't really do anything." And didn't really realize that the leaps that I had taken and what I was able to do, from when I started in the position to when I was finishing in it, was not anything that could be matched by a normal, yearly salary, comp rates, whatever.

Avi Freedman: With the currently active job market and the sustained bubble, hopefully that will be less of a problem for people as there's a lot of sustained recruiting going on. But, that said, a lot of app people think networking is just APIs and magic and you know, no one needs to know all the underneath. So it is networking type companies who are there. And definitely finding ways to get involved in communities is a good way to build those networks where you can find out what the interesting opportunities are and make sure that you stay connected there. So I also, it wasn't intentional, got lucky falling into NANOG and networking communities, but not everybody is as social that way. So with virtual, it'd be interesting to see how we, in the after times, evolve that and try to be thoughtful about how do we take that tribal knowledge and help people get into it, because it's tough.

Nina Bargisen: It's tough. It's been a tough couple of years. I'm saying a couple of years. It's almost, I think it's going to be two years, at least, before... Like most of my conferences this fall is turned virtual already. So there's going to be a gap. And also the new people who have joined the community in these years, we don't know them, we haven't met them. Funny thing is at RIPE, some of the feedback from the virtual conference is that there are some people who are more comfortable with that format. So they feel more included than they would when they go to physical conference.

Avi Freedman: Interesting I'd love to... You can send me a link after, we could talk after about that. I'm curious about that because the lobby track or the bar track of where people actually say," Oh yes. This vendor. This bug. That person." The stuff that you're not going to say-

Nina Bargisen: You're totally missing that one. Yeah.

Avi Freedman: Yeah. And I go to the things at a NANOG if I want to ask questions, but generally don't sit there, clickity, clickity, both because of meetings with customers, but also just because I run into people and you know a lot of people. I'm sure 400 people that you could run into and talk to. And thanks for calling out some of the stories. I think it's important that we all remember that even when we see the grand old women and grand old men of networking that, if someone is sitting next to you, to look for signs they want to engage and try to be welcoming. One of my great experiences was I was at NANOG and somewhere I have the Cheswick and Bellovin book, which was the first book on firewalls. And so Steve Bellovin was standing there and he was talking to someone about something and I was just sort of standing there. And then he just turned and engaged with me. He's like," My son thinks that I should be a better nerd and play poker like you do." It's like," Oh, he knows who I am." And we had a conversation and I wasn't going to bother him because he is like a great person of networking and all that. And so it's moments like that where you can meet people. And so that's been a lesson for me. Someone said," Avi, when you're talking with people sometimes, someone comes over and you sort of look and then people move and make space." How do you make sure that if you see people wanting to engage and get in that we can do that? And maybe that is easier online. So yeah, I'm curious if there's ways we can build that because there are people that say that the clue on the internet is constant, don't teach people because you know, they'll screw things up. I've never believed that and we'll die out if we don't get people in.

Nina Bargisen: Everybody who's curious about how shit works we should be told. You have to feed the curious. You can't make people who are not curious about it curious about it, so feed those who are that's crosstalk-

Avi Freedman: Yeah, that's a good way of putting it. And I think we can learn from the peering and interconnection world, which if you go to GPF, is more diverse than, as you said, the NANOG. You go into the people talking about MPLS iterations and, not to pick on any specific topics, but the more you're talking about vendors and routing protocols and stuff like that, it does get less diverse. Awesome. Well thank you for being so open and sharing and thank you for being a guest on Network AF and look forward to seeing some of that RIPE data and perhaps we'll talk again for an update in a few years.

Nina Bargisen: Awesome. I'd love to. Thank you.