alright now to start the program off with a real treat we have Dave Schaeffer the founder and CEO of cogent networks we've been looking for a talk like this for a long time it's a chance to hear from somebody that built a business from the ground up the business is the Internet and he's going to tell us why the Internet is the only network that matters so thank you Dave [Applause] thanks Shawn and I'd like to thank everyone for the opportunity to chat and thank Nanog for the opportunity what I would like to talk about today or maybe some bigger issues that we don't necessarily deal with every day and think about everyday you know the people in this room have a great responsibility we are effectively the plumbers for the Internet billions of people about 4.4 billion people regularly use the internet globally but few people could understand what the Internet really is and maybe more importantly what the value the internet creates for society and 18 years ago I had the opportunity to form a company from scratch and literally made the bet that the internet would be the only network that mattered and in 1999 that was a fairly controversial decision everybody believed data was going to overtake voice but the internet was never viewed as a network that was going to be ready for primetime that was going to cannibalize every other network in every other service in the telecommunications industry and the Internet has brought tremendous change we see the positive change in the way in which we conduct our lives the way in which our society functions but those of you who grew up in the telecommunications industry know that the Internet has also been responsible for financial distress you know Jack put up a slide of all those companies that were combining and that's just this year the reality is this is an industry that is under relentless pressure and our job is to do two things one make the internet work better but to make sure that the Internet is survivable and continue to deliver that value to society so Bob Medcalf who you know takes credit for inventing Ethernet coined a phrase that became known as mid-caps law the value of the network is equal to the square of the number of devices or people connected to the network that actually dramatically understates the value of the Internet the value of the Internet is really the square of the number of people connected times the percentage of their day that they spend on the Internet then multiplied by the value that time is creating for them and some of those inputs are easily measured the value one is difficult to measure but what the Internet has done has resulted in global GDP growth over the past twenty years of about one percent or a hundred basis points with the Great Recession that really says that virtually all of the economic growth around the world in some way can be traced back to advances on the Internet and what we do with the internet oftentimes appears to have little or no value but it is those few great ideas that truly create value and that value comes from literally millions of innovators who now have access to a global footprint and the barrier to entry to create new business models has fallen dramatically all of that has occurred because of what we do as Internet service providers or engineers supporting the internet we have to understand that the that is actually comprised of a series of fundamental components what we do is we produce interface routed bit miles connected to other networks if you're a salesperson selling transit or dedicated internet access the two ways in which individuals organizations connect to the internet if you told your customers you wanted to sell them interface routed bit miles they would think you're crazy but we as engineers need to relentlessly drive down those costs in each of those dimensions so the fundamental business building blocks of cogent when we put our business plan together were 5-1 that bandwidth is a commodity and the only way you can win is be the low-cost producer to local networks are a natural monopoly and as someone who's attempting to over build those networks you need to be very thoughtful about where you deploy capital in over building those networks third you needed to get the cost of production down to the very lowest point and that is the utilization of IP directly over DWDM protected at layer 3/4 to interface to the customer into other networks you needed to use Ethernet not because it was necessarily a better technology but it was ubiquitous and the cost per bit transferred over an Ethernet port was generally about three orders of magnitude less than the cost to move a bit across an SDH or telco interface and then finally what engineers often forget is we can build the very best work but if we don't have customers it's a science project not a business and it's not self-sufficient and self-funding so you needed to be able to have a value proposition that would lower the cost of revenue acquisition it is these five basic building blocks that allow a company to be an Internet service provider and not a telecom company now what's interesting is the Internet has some very positive attributes the first one is it was a network that really had no purpose that's really the first time in history that happened whether you think of a telegraph network a telephone network a cable TV network they all had a specific application in mind and the internet was different it was a public network not a private network and it was application agnostic the second key differentiator about the Internet is it was originally conceived as an overlay Network it was going to sit on top of infrastructure that already existed and as we think about those networks I'm going to break them into two constituent areas and then one has three technologies one has one the first is the backbone transport network and all of the original backbones were built by renting excess capacity from telephone companies basically wavelengths or SDH services across those backbones that technology was inefficient and irrelevant for the Internet the internet did not need the 50 milliseconds that a SONET Network would deliver it didn't need the Dax's and ATMs to front-end the class 4 and class 5 switches they were all involved in the migration of those bits from market to market so because of deregulation in the mid 90s there was a proliferation of fiber bills particularly in the backbone prior to that if you wanted to have a facilities based network the way you did it is you went out and for three or four years secured right-of-ways from Road systems from railways from pipelines you dug trenches you pulled conduit and cable and you then had a facilities based network because many companies did that before us we were in the fortunate position to be able to go out and buy dark fiber people laughed when I went into the room and said I want to buy a single strand not even a single pair I want to buy a national footprint with a single strand of fiber and what engineers did not fully comprehend was the nascent nature of wave division multiplexing and its ability to continue to advance at an exponential rate so we heard Jack mentioned the growth rate of the internet somewhere around 30 percent compounded for the past 26 years that's phenomenal there's actually no other industry in the history of the world that's had that sustained and that long of a growth rate at that high rate and that's in fact only the beginning we still have a long way to grow that growth is going to come both from new applications being developed specifically for the internet and from applications that exist on legacy networks being ported over to the Internet you know we had an introduction from Dave Temkin and his business is basically porting over the viewing of video content from a linear platform to one that is a multi-user multi dimensional platform that allows anyone to reach content virtually anywhere in the world when they want it I don't know our price point that is an explosion of choice and it's a reduction in cost all made possible because of the Internet that's an example of one application being moved voice is an application that is already migrated predominantly to the Internet and then finally the last bastion of the legacy telecom world has been the private VPN market that is now on the verge of being migrated to yet another over-the-top solution and these changes are possible because of those advances in wave division multiplexing so the first wave division systems for channel systems were deployed by actually Western Electric before it was loosened in the mid-1980s old for channel systems with 200 gigahertz spacing and systems that were supporting basically for concurrent oc12 doesn't sound like much capacity today the cost to move a bit a mile on existing fiber has fallen at a compounded rate for 32 consecutive years of 80 percent per year at the core of the internet that is what has transformed our industry and let me put that in perspective on the Internet today we carry somewhere around fifteen 1800 petabytes of information the Internet is a measurable thing it's big its amorphous but it is measurable you can measure how many bits go in how many bits go out how they are packaged and how far they go it's actually a number that we track on a routine basis the average bid on the public Internet is put into a packet size of just under nine hundred bits Ethernet supports 64 up to 1506 900 is about where the internet works the average bit on the public Internet travels about 2,700 miles and it can go one of two ways it can stay within a given service providers network or a controverse multiple a s's the average bit that's including those bits that stay on a network end up traversing approximately 2.4 a s's and they end up traversing about eight and a half routers from start to finish 2,700 miles 1600 petabytes but all that is possible because of that optical transmission core if we talk all of the traffic passed on the public Internet today with today's off-the-shelf technology that's basically 200 gigs per channel you can buy that equipment today 400 gig per channel equipment is being early field trial but 200 is shipping today and using both the C and L band you can support a hundred and sixty wavelengths at 50 gigahertz spacings those are the wavelengths that are optimally served in the EFD a band basically where erbium doped fiber amplifiers do their most efficient work now we can widen that band if needed we can go to higher modulation rates but the ability to drive down cost is coming from both the elimination of intermediate oh we owe optical to electrical to optical conversions when we first built our network which has approximately 58,000 miles of intercity to rust really we actually needed about 4000 conversion points in the network that's highly inefficient because each time you add an additional wavelength you have to add the transponder cards in both directions today we have been able to reduce that to only two locations globally where we're doing Oh EO conversion only for signal regeneration the router can actually perform multiple functions so rather than having the router be just that a router the router can also do the OEO retime reshape the signal it can also do the functions of taxing and multiplexing that normally had to be front-end so it's these reductions and cost in these efficiencies that have given us the first leg of our advances even with that 200 gig transport network and if that network was lit on all hundred and sixty wavelengths we would probably be only utilizing about 15% of that capacity to carry 100 percent of the world's traffic but it actually is even more stark than that there's over a thousand pairs of fibers between every major city pair in the developed world two thousand fibers we can do all of this whole one fiber not even one pair if you're running the network in a bidirectional mode now as it turns out most vendors have elected to go to uni directional systems and a to fibre solution is better than a single fiber system but there is no reason why that cost of transport at the backbone won't continue to fall at a comparable rate we have the ability to use more of the optical spectrum that EDF a band only utilizes about a percent of the spectrum and the optical waveguide of a piece of fiber so for all intents and purposes at the transport layer we can see costs continue to fall at this 80 percent compounded rate for the foreseeable future now let's go to the router in the router you've got a different set of economics routing is complicated it's open source based but it requires some very specialized processing the t cam and a router is somewhat unique and the market for routing equipment has been more concentrated than for transport equipment so the cost to route a bit has actually only fallen at a compounded rate of about 40% per year so transports falling twice as fast as routing in addition to that the interfaces between the transport network and the routing network have become much more streamlined virtually all the networks today utilize pluggable DWDM optics so when I scored a cogent I remember going to an optical vendor that's no longer in business that was leading the 10-gig revolution and the general manager of that business unit gave me 10-gig transponder and put it in my hands and said this cost 1.3 million dollars I've looked at him like it was crazy and I said you could buy a nice house for that and have a lot of money left over and he kept arguing the yields were low on the lasers the technology was complicated well the reality is it was just the production scale was too limited today that level of functionality exists on an SFP pluggable for under $50 so from 1.3 million to $50 is a pretty dramatic price reduction and it's even more profound than that because it's now a direct pluggable as opposed to needing a piece of transport equipment on the front end but in building the network we also needed some things that didn't really exist we actually wanted to build a fully routed Network well the problem was a router cost too much I know there's a lot of boxes that vendors call routers but they're really switches masquerading as routers they don't carry a full table they won't run BGP they won't do MPLS you can't turn net flow on them there's a variety of things those boxes couldn't do our network was built as a series of concentric rings for physical diversity roughly 710 Metro rings about 31,000 miles of fiber 198 markets and at the beginning we were forced to protect those local networks with a subset of the routing table we were running OSPF with a local neighborhood today we can run a full GP we can do IP fast reroute because every location on the network has a full router we operate about 6500 routers on our network we support about 18,000 active bgp sessions and support just under 60 100 unique AAS is connected to our network our network allows us to scale at that volume because it is completely routed all of the switch infrastructure boxes that had limited capabilities have been removed because that cost of routing has fallen low enough to make it affordable now let's talk about how and where to serve customers and serving customers your an over builder if you want your own network so today there are three existing networks that can all be used for the Internet one is a twisted pair telephone network it exists ubiquitously in the developed world Sprott of lee scarce in the less developed world but we have roughly 99% teledensity in the OECD or developed countries around the world that network has some positive attributes it's a hub-and-spoke network meaning everybody has a dedicated connection and the loop lengths are relatively short generally about 3,000 feet in North America about 1,500 feet in Europe in North America we have about 223 million pairs of twisted copper almost exclusively 24-gauge and it's always terminating at the endpoint and then on the other side it's terminating in 23,000 cos that infrastructure exists it would be great to leverage it the challenge is that the bandwidth on that twisted pair is limited even with the most aggressive vectored dsl solutions ADSL solutions we can get about 1 megahertz of bandwidth and because of Shannon's limit we can effectively only deliver across that probably somewhere between two and three hundred megabits in the ideal or theoretical case but in reality we deliver much less because there's low coils bridging caps and that copper plant is a hundred and twenty years old so at the end of the day even though the telephone companies had a bigger footprint they lost the end-user connectivity battle the second architecture that exists is hybrid fiber coaxial and it exists primarily to deliver one-way linear video it's a coax network that started life as analog it's generally been upgraded several times maybe to of 450 a 750 and 860 or 1 gigahertz digital system and it actually modulates at a higher level so rather than being simple binary on/off it uses qualm 256 256 discrete energy states of information transfer but as the number of information states increase the noise for increases relative to the transport and it becomes harder and harder to use DSP technology to pull useful information out you're also sharing that plant in two ways one you're sharing it with linear systems and secondly it's a party-line cable was never designed with unique customers in mind it was a shared bus in America the average cable plant has about 300 users concurrently sharing that bandwidth that means it works really well X up to peak hours when everybody's turning on unicast channels at the same time and throughputs drop dramatically so what carriers have done has bonded more channels together use better encryption technologies such as DOCSIS 3-1 and they have also reduced the number of linear channels devoting more channels but ultimately it's not an efficient Network either for delivering the Internet it's an intermediate network let's touch on the third Network the mobile phone network mobile phone networks were designed to carry analog 64 bit conversations that were voiced that typically had a nine point six kilobyte VOC order on them that's all your phone was equipped with in a 1g system and the cost to deliver those bits wirele is generally about two and a half to three orders of magnitude so almost a thousand times more expensive than fiber so we're describing a world where the internet sits on all three of these access networks and all three of these access networks are deficient what we need is ubiquitous fibre deployment but the challenge is no one can make money as an over builder Google's tried it Verizon has tried it business carriers have tried it and the reality is less than 20% of the u.s. this fiber connected now there are still builds going on we saw Jax map up here from Zao they're an aggressive builder we buy dark fiber from 222 different suppliers around the world we're a customer of that dark fiber not a builder of we then layer on this metro and long haul wave division multiplexing and we deploy that by connecting to the highest traffic locations that brings us to a target market we would love to be able to sell services to as large of an addressable market as possible every one of you that works for a service provider would like the same every one of you that works for an application provider would like to get to as many customers as possible so having a fiber footprint everywhere would be great the problem is no one can make money by overbuilding the two or three networks that exist today and the reason is your costs are proportional to homes or businesses past your revenue is only proportional to homes or businesses served and your cost of revenue acquisition is high making it difficult to get an adequate return on a speculative fiber overbuild we cream skim the market at cogent we sell services to end-users it's 62% of our revenues and we only sell to people located in skyscrapers so what does the average building we connect to look like it's 41 storeys tall has 51 discrete businesses in it and it's 12 and a half football fields on the roof anyone who's driven around the country knows that's not a typical building and it's the only way we can make money as being that selective about where we build a network we touch about 10% of all office space in North America about six about eight hundred and seventy million square feet out of eight point nine billion square feet of office space that leaves all the single tenant buildings the factories the hospitals fools and the homes 130 million homes in North America is unprofitable to serve that's how we built a dia business for end-user customers the second thing we have done is focused on selling transit to the people in this room that's probably how you know cogent that tends to be a controversial business we aggressively try to price our service to give people the very best value we directly connect in 43 countries to 980 third-party carrier neutral data centers so we've heard the gentleman from any clinics talk about how they went from 15 to almost a hundred and twenty data centers we're in 980 and adding about another 80 or 90 a year around the world and in those facilities they are supermarkets for bandwidth we are selling an undifferentiated fungible commodity to a sophisticated buyer those buyers want quality equally they want price and to succeed in that business scale matters so telecom services has another unique attribute that many other industries don't have marginal cost is always below average cost that says the incremental customer costs you less than your installed base in economics there's a concept of a backward bending supply curve at some point you hit a threshold where you've maxed out your production capacity and you have to do a step function in costs raising your costs in the short term to then ride the cost curve down again that doesn't happen in telecom and for that reason scale does match as scale comes with price and efficiency so for 18 years we have been relentless in driving down the cost to produce interface routed bit miles taking advantage of every one of those attributes that I've talked about in the market and in technology now for all the good things the Internet has it actually has some pretty big structural problems as well that people don't dwell the first of those is an inherent free-rider problem listen we were accused of it for a long time I'll be honest I worked very diligently to make sure that was not the case but many of our opponents came to us and said cogent is dumping traffic on us trying to use our network get a free ride we have always from day one agreed to offer meds to anyone who exchange meds with us - always furthest exit route we today sell to about 60 100 other service providers around the world I can tell you that the conversation with every one of those 60 100 started by them saying we want to be your peer we want free connectivity and only want cogent routes we today do peer and that's a controversial topic we only have about 30 peers globally we have an objective set of standards based on network scale network size network traffic and number of AAS is connected we peer with anyone who meets those criteria independent of traffic ratios oftentimes last mile network heavy networks have said ratios are the way in which we agree who we're going to connect to know it is traffic so what will three became popular for popularizing the concept of bit mile peering it's kind of interesting we propose that to them nearly a decade before they came out in support of it today that's viewed as a reasonable way to ensure networks carry their own load so the free rider problem everybody trying to get a free ride everybody trying to use their market power actually becomes anti-competitive and becomes detrimental to the internet ecosystem may not be that way for the individual getting the benefit but for the whole ecosystem it's not building out the infrastructure needed by the way the same is true of the opposite side of the relationship the large content companies so if you're small and you don't have very valuable content you fully expect to pay for transit if you're very large you say my content is so important you as a transit provider should give me a free ride I'm not going to build a network I want to ride and this is often masked by the deployment of CD ends CDNs are nothing more than distributed hosting companies if they're a third party CDN they are blending the combination of space power transport routing and interconnection with storage and processing into a finished product if they are a content provider building their own CDN they're doing the exact same thing except they're doing it only for themselves also there is a belief that Direct Connect is maybe an alternative or better for the internet and the Deaf the Internet has been predicted probably a 9 times over coach ins 18 year life and people said well coach ins going to go out of business because it's a one-trick pony focused on the internet and the Internet's going to be replaced by CBN's it's going to be replaced by cross connects and a carry or neutral it's going to be replaced by Direct Connect and there is a place for each of those technologies data centers that specialize in colocation and cross connect add a valuable component to the ecosystem some of them do that very cost-effectively others attempt to extract a monopoly profit and when they do that often times there's an adjacent Center right next door that opens up and supplant that direct connects between networks are far more efficient than shared exchanges with intermediate failure points and higher costs associated with them they're much more efficient than the old directly slowing connectivity model that would have never scaled but a data center operator has three different levers to generate revenue space power and cross connects and each one of them has their own model on how they kind of play with each of those three and we as network operators just want to interconnect we want interconnect to one another's networks and we want to interconnect to our customers in our case as a transit provider and when we do that we are looking for the most ubiquitous reach and the lowest cost and what is an interesting phenomena that's really more empirical than necessarily theoretically derived is that bit travel distances have not materially changed in eighteen years and at first blush that seems counterintuitive with the proliferation of content but it has been supplanted by the globalization of the Internet in 2000 you go back to the early and a dog's 85% of Internet traffic globally originated or terminated in the United States today that's down to the mid 30s about 35% and at some point we're only five percent of the world's population we're only going to be five percent of the traffic that's a ways off but we're going to get there and that's ok the Internet is a global phenomenon and can add value to everyone but those companies that seek to use their market power and not participate in the ecosystem Direct Connect from content to direct connect to a network operator if it's cheaper or better they should do it if it's done because a network is using its monopolistic power either on the content side or on the application side that is detrimental to this ecosystem so it is the free rider problem and the attempt to use market power to circumvent the engineering forces that are driving down pricing an interesting study that we commissioned we were very vocal opponents in the net neutrality debate and in the testimony around Time Warner's proposed acquisition by Comcast and in that discussion we found out that the average cost per bit for an end user for the past 20 years has only fallen at 1/8 the rate that it has for transit providers technologies are very similar yes there are some differences in the architecture we've talked about those but the 8 to 1 differential was really an attempt by those local access monopolies to extract profits to subsidize their legacy businesses that were being cannibalized by the internet the internet will be the only network that matters and the reason is it's better there it's more ubiquitous it's cheaper and it is completely flexible the applications are only limited by the 7.3 billion people in a world and the ideas that they have most of which will fail and that's perfectly acceptable but it goes back to that original concept of value creation that if a few of those applications catch people's attention end up having them use the internet more hours a day consume more bits per hour and more of them do it and most importantly they get more value out of the application you can't determine that telemedicine is more valuable than watching the next you know episode of house of cards that's an individual decision people should be free to make whatever decision and run whatever application they want and just like any part of our society there can be bad people there could be people that do evil and bad things with our technology we are not police we are not sensors we are not here to limit people's choice governments can do that and we have to adhere to those governmental dictates we've taken some heat from some of the people in this room for turning off certain sites when a legitimate government issued a legitimate court order to do so whether it was in the Arab Spring or whether it be in certain terrorists and senses it is not our place to be the policeman I don't condone I don't support those misuses of the internet but it is not our place as engineers I mean our tag line kind of says it smart people buy dumb pipes it used to be a joke every Network wanted to tell you we're going to move up the value chain we're gonna add value added services we're going to do something that other networks can't do we're gonna have better customer care and a reality is no company's good at everything no company creates all of the applications we need to be good at what we do which is running networks and we need to make those networks more reliable we need to make them more scalable and most of all we need to make them cheaper I often get asked the question how can you have a transit business when the average price per megabit in your short company's history has gone from a market price of $300 a megabit to a market price of $1 a megabit and for cogent or price turn at the same time has come from $10 a megabit down to below 10 cents a megabit in some instances yet we can make money and what people don't get is the underlying technologies are improving and you need to have the standardization of products before I came up here I was talking to someone who has product management responsibilities of one of my competitors and they said well we have thousands of products and we can't really make money doing what you do and I said that's what competition is all about that's okay we don't want to do what you do we have no core competency in those areas we have a core competency in delivering those transit bets globally to smaller networks and to content providers if you look at our transit business it accounts for 95% of the bit load on the coaching network 38% of the revenues it comes from just about 10,000 customers 6000 access networks about 4,000 content producing companies another common misconception is that were somehow traffic asymmetric in fact about 60 to 63 percent of our traffic stays on our network never gets exchanged and we disclose that to both customers we oftentimes say well you're double-dipping you're getting paid on both sides well that's true but we disclose that we show you the specific routes we give you the net flow data will show you where the traffic's entering where it's exiting and to what I asked with what capacity I don't know how much more transparent we can be and that's part of why we can drive down pricing it's that end user business that bothers me more because we can only do it in certain places we're running out of buildings that fit our end user requirements in terms of size number of businesses and aggregate demand because we so know value-added services that means we need to be efficient at the network layer those dumb pipes have to at least pay the bills so if we look at the total telecom market globally in the past 20 years there's been nearly two trillion dollars of investor capital private capital around the world that's disappeared in telecom telecom went from being just under 9 percent of the sp500 to 2 and a half percent of the S&P 500 many of the companies went out of business much in the capital that was deployed is now fallow and was wasted and that's all because of this level of efficiency that the Internet has wrought and those that work in integrated companies that sell legacy services along with internet products are now on the cusp of facing yet another huge challenge as we see the last bastion of high revenue per bit traffic MPLS VPNs starting to be supplanted with VPLS and SD when those changes are inevitable those products are delivered over a common Internet core and with that the cost per interface route a bid mile tends to be 15 to 20 times less expensive economics may sound boring to engineers but it's what pays the bills and what allows us to do what we want to do it's what allows us to create a network and build the scale and deliver the services we want what I'd like to do in the next 10 minutes or so is maybe shut up from a monologue and try to open it up and see if we have some questions from the floor [Applause] hi Dave Matt PTAC from Yahoo or oath or whatever we are today very good talk thank you one thing that as you're getting to the end there you were hitting on a topic that Jeff Houston had actually talked about a few Nanog which is if you're the person delivering the pipes you're essentially the the sewage company you are a utility provider your ability to extract revenue from that is essentially zero and if your business model is predicated on that you're in trouble how would you respond to that your with your business model saying hey it's okay to be just the utility that the provider of just the pipe do you think Jeff was wrong do you think there is value to be extracted from that or is it a race towards the bottom so oftentimes I get asked on a price per megabit this is a race to zero and no one is going to win second point is capitalism is the greatest system of economic allocation ever devised but in a capitalist system the system ends with no profits marginal cost equals marginal revenue systems are perfectly resilient and perfectly efficient businesses are always trying to create inefficiencies barriers to entry and the reality is any utility is a natural monopoly whether it be water or gas or roads or pipelines or electricity or telephone or internet or sewer and it was interesting you picked the one that comes up with a negative connotation a sewer and you didn't pick the positive one which is water I'll blame Jeff for that because that was his analogy you know running a local network is a monopoly it's a natural monopoly and therefore it needs to be regulated as such it needs to be built with private capital now public capital a rate structure needs to be put in place that ensures an adequate rate of return on that capital with an appropriate balance sheet of a combination of debt and equity those are all very good things now on the bad side monopolists always walk to do evil and I know Google slogan is do no evil but they do have 90% market share at least in a market and I remember meeting John Malone only one time in my life and he said something that stuck with me he said monopolies are great things the problem is I can't find enough of them and that's absolutely true I've worn a monopoly every businessman dreams at night of going to sleep and waking up and having no competitors well guess what if you build a local network you shouldn't have a competitor and the reason is that competitor ends up destroying value because he's building duplicate of networks that are going to be underutilized he's raising the goods for everybody and building that now to protect people you need structural separation network and operating companies need to be separated almost the exact opposite of oak your current employer who's looking for a vertically integrated stack and a structurally separated business was tried in a telephone era and eventually was defeated in Pennsylvania by the Pennsylvania Supreme Court and ultimately the internet did what legislation couldn't do the internet passed structurally separated the application from the network and has caused people it caused them a problem at the exact time that they have other problems so they have pension obligations they have debt obligations dividend obligations and they're not able to meet all of those obligations so usually the easiest thing to cut is people so I don't know what the exact headcount at Yahoo is but I think it's less than it was before and my guess is we come back to now 74 or 75 it'll be lower again so I would propose that regulation of least in North America is ass backwards it's tried to create competition at the service level and it has not addressed the need for private capital at the network level so you have to wake up with yourself in the mirror and say what am i and a local network as a monopoly and therefore it needs to be regulated but it also needs on the counter to have an obligation of universal service and quality of service so there's no shame in delivering roads or electricity or water there are many aspects of a civilized society that are commodities we can't live without any of those in some countries it's viewed the Internet is necessary for life I'm not sure we're here yet but I think 10 or 15 years from now that will be viewed as as critical to life as any other utility you get so the answer is get used to it you're a dumb pipe and just live with it whose guy now that give the quick I thank you for the answer on that okay Center microphone okay hi my name is Daniel countback I'm an internet citizen thank you very much for the nice talk thank you above all for not using PowerPoint or anything and just talking coherently I really like that also thank you for talking economics to engineers that's much needed anyway I'd like to address the last mile thing to them just give you an example of where I live I live in the Netherlands I live in the smallest village there used to be the even the smallest incorporated place yet I have two strands of fiber in my home why is this is it's because basically that bit is not run by the government it's not run by a benevolent company or organization it's run by some people who have who are over building but they have a longer outlook on when they make money it doesn't have to be the next quarter it can be ten years can can be 15 years it's more like real estate and I think that's a little bit that we're going to need if we really want the Internet out of the overlay networks and get to the point where you can sell maybe me some service so yes regulation is one but the other is basically getting it into a sector of service sector of business that has a longer outlook you know that looks about building infrastructure and not bigger taco was you know to turn of turn in a profit in the next quarter do you think that's a way of maybe attacking us so I am a big believer in private enterprise there are certain things that are necessarily social goods some of those are military defence or certain kinds of health care that we as a society have to take care of everyone that's part of being human in a ah world I'm not sure the internet hits that criteria I also am actually a bigger believer in people having a longer term view than you are I'm not a cynical cynical of capitalism being totally quarterly driven I'm a CEO of a public company I am measuring every quarter every day there's a voting machine my stock goes up goes down shareholders get happy they get angry and the reality is I think about things over a longer timeframe in the long run markets are rational we have a huge actual problem in a world that there's a lack of growth because of a lack of growth most the people of our age have a real problem in expecting to retire at the standard of living and having their pensions fulfilled the problem is not unique to us it's everywhere in the developed world and part of it is this short-sighted nature of capital as opposed to a long-term view there is plenty of capital available looking for good rates of return and that capital is willing to take a 10 15 or 20 year view of the world even 30 and 40 years we've seen countries like Italy issue hundred year bonds where investors are not expecting a principal to repay for a hundred years that's beyond virtually everyone's life time particularly anyone who's buying the bond there's got to be at least 10 years old so very very few people live to that age so obviously there's people that take a long term view but back to investment the reason people won't invest and build fibre to your village is one of two reasons one there's certain places where the density is just too low to create an economic return and the only way you can get that type of build in those locations is through cross subsidization people in big cities people pay for people in rural areas in our industry people in business pay for residential they pay higher rates to help subsidize and those subsidies are actually okay the other thing is that you have to be convinced there's not going to be an over builder so the problem is you come into the market and you build a new fiber round you give this great service to someone in the village in the Netherlands and 20% of the people take the service 70 or 80 percent of the people stay on the old services it's good enough for them that maybe they're old maybe they don't want to change maybe you don't want to watch Netflix they want to watch you know you pee you PC services that are linearly delivered in Netherlands the old ziggo Network and they're very continent you said I'm not switching so you know all of a sudden of deploy capital to pass everybody but you only got revenue from 20% of the market and you've got no protection that another over builders not going to come along capital is not going to invest in that environment so you need to create a regulatory framework that says capital is invested it's protected and it's managed to an adequate return and that return can be high enough that will incent pension funds to flow money in build it with private capital yes people will get dividends evil capitalists will make money but the society will benefit by privately constructed infrastructure we don't have those regulations today maybe they'll exist someday they actually exist in some other industries in the electric industry where there hasn't been this level of technological tumult it still does exist I've got a couple more also just to know we have a hard stop in about nine minutes so if there is anybody that has a question that isn't at a microphone please line up so we can manage the queues and Owen at the stage right is next Owen DeLong Akamai Technologies you talk about natural monopolies and I think one of our biggest mistakes that we've made so far is allowing the natural monopolies that exists at layer 1 to be leveraged to a monopoly up through at least layer 3 and sometimes beyond that and I wonder what your thoughts are on possibly putting a regulatory framework in place that prevents layer 1 monopolies from being leveraged in that way and requires a availability of those layer 1 capabilities to everybody that wants to provide layer 2 and above services I'm actually go go down to layer 0 the physical layer as opposed to even the service layer of transport I think there should be complete structural separation so part of the bargain in getting that guaranteed rate of return is that you cannot sell services either yourself or through an affiliated company you have to make your infrastructure available to all on a unique and on a competitive basis that doesn't advantage one company over another so I envision a world where at the infrastructure level we may have a Facebook ISP and they may go to every customer and say you know if you use 10 hours a month of Facebook time you get your internet service for free you may get Google saying well we're gonna one-up them and all you only need to use Google search 5 hours a month and you get free internet and it may even go so far as someone will say you know if you use this 10 hours a month we'll pay you to use the network and we'll pick up the cost that's what a market is about and you know if Verizon where if Comcast says we don't want to be in a network business anymore that's fine spin that out you could be another service provider maybe you're competing directly with Netflix you're a curator of content and an aggregator and a markup of content that's perfectly acceptable but the whole idea of applications and networks being bound together in a vertically integrated one through three or even one through seven stack ends up creating this perverse incentive and ultimately hurts the system hi there my name is Kevin from eBay actually had this question more as an individual however sort of curious is the intersection of all of the stuff we've been talked about in the Q&A from the capitalist side to the stuff you're talking about with you know vertical integration and all that stuff how do you view from your perspective the sort of current environment where you see infrastructure providers starting to capitalize the actual data and you know looking at what's actually flowing across our pipes sneaking into it you know learning about who's doing what on their network and starting to sell that data and sort of both the capitalistic and regulatory framework so there's historically been a concept that a common carrier which was subject to regulation would not be able to look into the content that was being carried and was not responsible for the content that basic concept was what was used in net neutrality when the FCC attempted to support the idea of net neutrality no blocking no throttling nope a prioritization adequate interconnection the only way they could do that was classify ISPs as a common carrier the dirty word title to and that then said you were precluded from being in this kind of intercept and intelligence business many companies have built great business models by giving away a bunch of free services to people in exchange for the customer paying for something that nobody else would buy from them their privacy so when you go on Google you go on Facebook these are great services they're free most of the Internet is dominated by free services and the way people pay for those services is with their privacy with their information and I think that should be with knowledge and it should be consensual I don't think service providers should be in the business of extracting that we are dumb pipe companies we should be at the layer zero layer one level and not getting into the application layer so I know there's an attempt to waive the privacy restrictions on service providers and they're very subtle about the way they weave that argument together one is the customer information by billing information my residents that's probably legitimate you can't deliver the service without having that and it's information you have to give to a service provider that's very different than saying I watched a certain show at a certain time or I bought a certain set of products and now I'm going to monetize that information that's a very different use of personal information and I am more in favor of this separated model and I know a cogent while we carry a huge percentage of the world's traffic we are scrupulous of not looking at what's inside of that all right thank you very much Thanks well I think we've got there's any other questions if not I know people are probably going to want to take a break here in a little bit but I want to thank everyone for listening I know this was maybe a little bit of a different presentation than many of the Nano presentations which tend to be very technical and necessary for our industry and I applaud the work of nano and other engineering conclaves that allows us to move the profession forward but I thought it would be helpful for me to explain how you can actually build a viable business because without a business none of us would be here there'd be a much smaller room of people that work for research institutions but almost everybody in this room has a commercial reason for being here a company that has a profit motive and while many of those Bodum's individually may not be conducive to the internet when they're all put together in an ecosystem they can be and I really think we need to get away from the mindset that the gentleman who asked the question earlier that we're selling sewer pipes we're not we're selling precious precious water if you're in a desert you need water to survive you can go a week without food you can go only about three days without water I would argue in a connected world there are many people can't go three days without connecting to the internet and my guess is in the future that's going to be even much more integrated into our lives everything we do every aspect of our lives has been touched by the efficiencies that the Internet has wrought we should feel proud as people who participate in this industry not ashamed not as the roto-rooter ban on the sewer but rather is someone who brings a vital service that's promoting economic growth and helping lift everybody in the world to a higher standard of living thank you all very very much for the time [Applause]
