APNIC Technical session 2

Show video

foreign [Music] [Music] [Music] foreign [Music] [Music] [Music] and yeah [Music] thank you [Music] exactly yes that is starting and I we want to just start next session next session is at technical session two and if and and next decision technique sister to Welcome to the uh this session and we have uh my name is chair I listening to oxy apparently and I'm working at the program Committee Member of the epinic 50 shakes and we have a city speaker and the first social speaker is a huge spider University of the okra breathe all right sir thank you okay um so I'm going to talk a little bit about starlink today so uh Starling of course being the uh um the internet service that sort of started up about well now it's really almost one and a half years ago um so it's the service that um Elon musk's SpaceX provides and so this is work that I've done together with my PhD student Wayne Raya um and uh we got ourselves a styling units um sort of in November last year and sort of started playing with it a little bit started doing measurements and um these are the sort of insights that we've sort of since gained now can I just have a quick show of hands who of you has used starlink there's a few of you okay wonderful great wolf you found it a bit of a mystery for whom did it work okay cool great so anyway let's talk a little bit about it so what we see about starlink um so this is from a website called sell my cell live map.space and um one of the things about starlink is all of those white dots are satellites and on this um sort of opening map you can kind of see um why it's a little bit special so if you go to that website what actually sees you see those white dots moving so every white dot is a satellite those chains of white dots that you see there are recent launches of satellites that aren't quite at station yet and one of the things you also notice when you look at that world map you notice that most of the satellites are concentrated um in a region between 53 degrees south and 53 degrees north uh with only a relatively small number over the over the poles and there's a reason for that and that reason is that those satellites are basically or most of those satellites are orbiting at an inclination of roundabout 53 degrees which means means they're literally sort of going from 53 degrees north in a South uh easterly Direction then 53 degrees south at some point and then go back up again to 53 North um so if you look at that map live then what you see is you see them sort of kind of moving diagonally a small number uh in sort of neapolar orbits so they'll basically be moving straight north to south so um the other thing that that brings with it is that you'll notice that the highest concentration of those satellites is close to those 53 degrees uh you know parallels and that shall play a little bit of a role as we go on through my talk okay right so first of all what is starlink uh well I just explained that this is the SpaceX internet service so basically what you uh do get for this it's a so-called low earth orbit surface low earth orbit means that the satellites are in the vicinity of about 550 kilometers above the surface of the Earth which is much much closer than the geostationary satellites that I bought 36 000 kilometers away from the equator that people used to work with when they used to work with satellites and that some of you probably still work with um and that makes for much much lower latencies to start with but it also makes for a much much nicer link budget because the satellite transmitters power with spherical spreading doesn't actually reduce quite by as much by the time that it reaches the user terminal on the ground and similarly when a user terminal transmits you don't need those huge antennas anymore okay so the whole thing basically also comes at a relatively uh nice cost so your um initial outlay is distinctively this side of a thousand US Dollars and um a monthly subscription ballpark depending on where you are costs you in the ballpark of about 100 uh US dollars so this is roughly what most of us would pay uh for a home internet connection so this is the simple story and it gets quite complex from there uh by the way in case you were wondering where I've got all those pictures from from um I mean this is a map from Google Earth obviously but those those other ones that look a little bit sort of futuristic and like taken from space um the artificial intelligences take on what starlink might look like so um don't uh you know don't take that um for what it's really like okay so basically how Ling does that connecting to the internet for you so you get um a so-called dishy these days that's a rectangular terminal um roundabout the size of an A3 sheet of paper and um when we have this in the southern hemisphere um that'll basically rotate itself um we'll try and find the highest density of satellites in the sky which in the southern hemisphere in the sort of moderate latitudes uh like we have them in New Zealand um it'll basically face Souls towards the 50 Foot a parallel so uh what then happens is basically say for example you sit somewhere near uh near the escape in New Zealand and you have one of those units um it'll basically try and find itself one of the satellites that it can see a satellite that does have capacity to carry its traffic and it'll send that traffic via that satellite to a Gateway now the starlink gateways um starlink doesn't tell you where they are however when you want to know what starlink is doing a great source are regulatory filings with the FCC or in the case of New Zealand with the ministry of business Innovation and Enterprise hello guys MB I want you guys to fund me next year hopefully maybe maybe not I don't know but basically their radio frequency spectrum management division they publish or licenses that radio Services hold in New Zealand and so Starling holds a few licenses and from those licenses we know where their ground stations are and one of those is in Cleveland right next to Auckland which is where I live and um that's where you basically Connect into the uh fiber Network and you can sort of see them from the air quite easily they're typically nine little right ohms typically about a meter and a half ballpark in diameter I saw that so they're quite compact okay so what happens is that um you know all low earth orbit satellites do actually move quite quickly so after a while those satellites will move over over an area where you will have a slightly different uh set of users connecting to it and where maybe it's no longer optimal um for your dishes to be communicating to that through that to a Gateway and uh in that case it'll switch to another of the satellites that it can see so one of the common misconceptions about starlink is that when you had low earth orbit satellites in the past there were usually either individual satellites or um they were sort of very very small constellations so the idea is you would track the satellite as it went over and you track it for as long as you could um because that was the only satellite that'd be coming past for a long time um but these days uh with starlink you can actually typically see um and sort of areas like like here or New Zealand you can typically typically see about a dozen satellites up there and so basically what happens is if you can see a dozen satellites you can actually switch rather quickly between them and we see the moment's time that that's actually what's happening so um so we get all those handovers and so the typical period between the handovers is under a minute okay so the other way in which you can connect get connected to the internet um you know via you know via starlink is if you're so far away that there's no satellite that both a ground station and your Dishi can see at the same time so in this case you'll be using um that half of the styling satellites that use inter-satellite links and those are laser links in space and space you can use laser links with impunity because there's no Cloud so there's nothing to to stop your laser and so the ideas you Uplink uh you know or downlink to a satellite that you can see that satellite then uses a bunch of laser links that take you to a satellite that can see one of the gateways and that then connects you to the internet from there so that also obviously needs handovers and generally quite a few more than your your entirely uh you know radio frequency based connection will um but obviously the rate frequency based connections or so-called Bend pipe connections they're only available if you're actually um you know close enough to a Gateway so this typically means you need to live within a few hundred kilometers of a gateway so we know that the laser links already work because um we know of a number of users in in tarawan kiribus that have tried this out for us and they get connectivity and there is no place anywhere near kurobas where you could be putting a um according a Gateway that would be able to handle band pipe traffic so in this case we know that the laser links are operational okay so the other thing that you um inevitably get if you Google styling on the Internet is people going but well what speed do I get now of course most of you are sort of from a bit of a technical background and most of you like me have probably grown up with the term data rate not speed um or bitrate or baud rate or whatever but you don't use the word speed there's only one speed really that we're talking about here that's the speed of light um unfortunately uh popular opinion has taken over so speed nowadays means megabits per second or gigabits per second or whatever so when you do sort of speed test measurements then this is kind of the sort of stuff you tend to get so you know for the kind of RV type subscription that we've got for our unit you get roundabout just under 200 megabits per second on average but it varies considerably so it sort of starts going down you know into the you know sort of mid to low uh double digits and sometimes it goes up to you know almost 400 or even just or sometimes in over 400 megabits per second and so you ask yourself well why would it vary that enormously right and um it turns out there's actually a reason for that and we'll talk about that in a moment okay so the other thing that you get um is that you can't get styling everywhere so availability is rather selective and there's a number of reasons for white selective so I'll just go through a number of those um regulatory issues so some countries simply don't permit starlink okay try and get it in China and and you know in India and various other places in other places like for example the Southeastern uh us you find that competing Services competing satellite services have gotten hold of the frequency allocations first soil starlink isn't available everywhere the Earth simply because they've run out of capacity um with a little uh terms of bandwidth they've got available over there so then there's other reasons and they're a bit more interesting um you know from the perspective of what speed do we get so if you look typically in the vicinity of larger cities or in the case of Tokyo for example also the inner city you find that you can't get styling there and so the question is why so if you look and and if you look at these maps online you'll probably find they've all changed a bit the reason why they're changed is because starlink basically the solar has a race going on it's got they've got a race going on between um between availability um of um sort of you know slots on the satellites that can be seen from there and the number of uh styling terminals on the ground and there's more evidence so those those those non-available spots that you can see though so those those cells They do change a bit over time right but if you look where they are it's quite interesting so for example um in New Zealand all of a sudden you know one day I got up a few months ago and Starling had lowered its prices previously they would ask for over a thousand New Zealand dollars for um for a terminal and then it said 199 dollars so price had gone down to 20 of what that originally asked for and looking at the small print it says in rural areas so nowadays they've put the rural area price up again by 50 now all inflation is bad but it's not so bad right so at the same time uh you'll find that the hardware price for uh for elsewhere hasn't changed that much it actually changed again last week I think we're now under 700 uh for for everybody else and 299 for people in rural areas but the interesting thing is what qualifies as rural for starting so for example um the city of Hamilton at Central business district so Hamilton is a city about 200 kilometers 150 kilometers south of Auckland and it sits in the middle of a big big rural area and Hamilton you know even when you're right in the city center you are Rural right now tell this to somebody from Hamilton that's an insult no in late August they stopped calling Hamilton rural but they were still only charging the rural price so obviously if this is rural you should be able to get it in most of Auckland because most of Auckland um is more built up than this so let's check this out so let's go to the Auckland lifestyle block this is uh just sort of outside Auckland I know somebody who lives just on down the road there and uh who recently got themselves starlink and what do you pay here you pay full price right this is Urban okay does anyone see any signs of this being Urban here I don't okay I can tell you why this is happening so basically why this is happening is because this is a rural area where the fiber doesn't reach so basically the New Zealand's uh New Zealand government's Broadband project hasn't gotten uh you know out to there yet so there's lots and lots of starlink users that are on private roads that haven't got fiber yet but those are people who live in town and who need a good internet connection in order to be able to work from home so they all went out and bought themselves styling so what's actually happened is that starlink has now reached saturation out there and so what they're trying to do is they're trying to discourage people from getting themselves starting so this is why you pay full price here uh for the hardware if you're in a place like Hamilton where the demand hasn't been that high because you get um you know basically fiber there and fiber is a better deal for most people then um you get a much much uh you know much much uh you know better deal there but if you're rural in the Urban Fringe then for practical purposes this is what you get and if I go a couple of slides back um then you see exactly that the sports where this is happening is literally the areas around um the sort of uh big cities where you get those lifestyle blocks um where people live but where the fiber from those places doesn't quite reach so okay so um we also started tracking satellites a little bit so in this case what I apparently have to do is start this so this is if you wish if you looked into your dishy and you thought your Dishi was a mirror that's that's sort of mirroring the sky for you and then imagine for a moment that you could see his dots on here the satellite that this is currently talking to so this is actually running at an accelerated speed but what you can see every time you're plotting a track here this is literally um uh this is literally a uh a particular satellite that's um Starling is currently talking to or your styling dishes currently talking to and then the next thing is basically typically between about 15 and 90 seconds later typically round about 30 45 second mark it'll switch to a different satellite and um if you fast forward this a little bit for me please hello oh done it okay so what you see this builds it up over time after boot up and what you see is that those all those black traces are satellites it's been able to talk to and the red traces are satellites it's not been able to talk to um generally because there's been some sort of an obstruction and in my case I actually ran this off my deck at home and that red bit that you're seeing is actually the corner of my house um and um the other rate bits you see are either dysfunctional satellites or possibly airplanes approaching um Auckland Airport obstructing the path I don't know exactly what it is okay so the other thing that people always complain about when it comes to styling is latencies because people can't understand why something that should have a physical path we're onto per path of under under 15 milliseconds is coming up with latencies um you know in the you know many dozens or even you know sometimes well over 100 into the hundreds of milliseconds of round trip time so when we do a trace route and these are Trace routes that I've actually done during Cyclone Gabrielle earlier in the year this was the big Cyclone that swamped half of New Zealand and um basically took a large part of our Communications out with starting being one of the things that still worked um and what's quite interesting uh that is that actually during that Cyclone we saw the round trip times going down so basically we've got better performance out of starlink in terms of round trip times that we got during normal operation and after the storm it's sort of all improved again but the interesting thing is you notice these are step changes right so this isn't it gradual um decline in round trip times this is literally sort of immediate drop and round trip times and my going Theory at the moment is that this is because those very lifestyle block belts around our uh around Auckland and some of the other places um where most of our styling users sit in New Zealand they had power outages so they went offline and um as a result of that the load on the satellites reduced um uh the queues are reduced in size and that's what's basically uh giving us the lower entry types yeah so and you can also see this when you're when you're doing a trace route uh through us through a starlink network from a Dishi you basically see that most of um the rtt is actually accrued within the starlink network and then beyond the starlink network to a server that's nearby to the Handover Point um it's basically bugger all Okay so uh talking about space lasers being used so this is places like kirobas so this is a trace um that we've done from kurobus and interestingly enough if you go from caribus no meta where in the world you're trying to talk to um in this case we're trying to talk to um you know a DNS server in Japan so we're Tracer routing to there and interestingly enough it always comes out in Auckland so everything goes through the Auckland peering exchange um or um through Hurricane Electric Wireless in Auckland depending on opposite you know which off the ground stations that it comes in through and where it's going but basically all traffic from kurobas gets taken to Auckland first and then it gets taken to the rest of the world where it needs to go so that to a certain extent uh ports paid to the theory that a lot of people had oh you know once we're going to have all those space lasers in place we're going to have really really low latencies from A to B and it's going to beat Fiber Well um sorry no it's not going to be fiber in a long time so talking about it beating fiber this is a statistic or the data that this is based on the statistic that the New Zealand government connect collects through its Commerce Commission and one of the other things you need to know before you get into styling is your speed advantage over DSL is largely in the download Direction so if it's it's data flowing to you if you have data flowing in the upper in the opposite direction it's you know not that great you're just marginally beating 4G um and if you look at the uh down up ratio this is the observed data rights average data rights that you're seeing over the over the 24 7 period or during Peak period you basically see that um that it's basically behaving much more like ADSL so very very little ability to send data up and yes I realize I'm running about out of time here okay so basically um give me fiber Max um you know any time of a starlink that's what I use at home and this is what where you know basically uh connect to a network drive at University that I uh put my uh lecture recordings on and so I couldn't do that with styling okay so then starting a few months ago came up with a sister this wonderful announcement and in fact some of the local telcos came up with the announcement saying in you know 20 to 24 we're gonna be providing a cell phone service everywhere in the country from Starling so question is how realistic is this and what are you likely to get now turns out the satellites needed for this aren't even in space yet um and even once they're in Space the problem that you're having is that your cell phone antennas are really really really small compared to the size of udishi and also they're trying to talk to what will be a base station that's literally you know hundreds and hundreds of kilometers away and there's a whole lot of problems that you're getting when you're trying to do this one of those is that you can't do the frequency you reuse that you normally do on a cell phone network so there's no what there's going to be no watching Netflix um you know over over a space connection like that however you'll probably be able to text and make limited um you know quality uh phone calls and you know putting the odd email out will probably be also an option but um in order to do that you should probably preferably go outside it's not going to work for from indoors like your phone normally would okay so my conclusion for all of this um a lot of what I've told you Noel may already be slightly outdated like those Maps were for example um they tend to evolve a wee little bit so starting's evolving really really fast it's working way better than geostationary and in some respects also better than medium earth orbit connections it's cheaper and faster in terms of what you're paying you know for the bandwidth that you're getting and a lot of people that you talk to who've so far not been able to get internet at all sites and absolute game changer for them it's a really really clever system as long as you're able to ignore what Elon Musk says on Twitter um and the question is is this going to get the 2.7 billion disconnected or underserved people connected um short answer we're into a few million users now so and this is probably still single digit number you know single digit million users and we're already having growing pains in the starting system so I don't think we're going to be seeing uh styling connecting you know two billion plus people anytime soon um so buffering and routing um for styling seem to be a construction site um and I wouldn't be surprised if we'd be seeing that evolving a little bit as well over the next Euro too and that's it for me thank you very much okay thank you very much presentation and maybe we can receive it with one or two questions from an audience and if we have a two mic uh sent over our room and if you have any question then [Music] um is there any questions okay please State uh your name Andrew from Bangladesh I have a question to that uh just just for an idea that uh what is the maximum capacity in a particular area Australia can deliver so that in case that how many people can connect in a particular area to satellite capacity is a really really good question so essentially as they add additional satellites um you're adding capacity um the problem that you get is that there's only so many satellites you can add until you get a satellite density whether she finds it difficult to distinguish between the adjacent satellites in which case you can't switch over to an immediate neighbor or you can't use the intermediate neighbor because another dishy that's transmitting to the neighbor of that satellite uh is already going to be using that frequency so this has gone up this is going to fill it up the other thing is uh obviously there's not not an unlimited amount of um you know frequency spectrum available so that's going to be a bottleneck and um obviously you get to compensate for that to a certain extent you get evolution in the technology on the satellites the satellites are going to become a bit larger uh probably once the uh you know the Starship launches are going to be available um so there's going to be a little bit of capacity lift uh but I think we're probably looking at a total yet to happen capacity growth of maybe a factor if we're lucky a hundred all up and so this is this is really rough estimate ballpark but we really need a capacity lift by a factor of about a thousand uh you know or over a thousand in order to get uh to the point where we could really reach everybody on the planet and also the other thing is this is assuming that the disconnected people on the planet are all equally distributed across the planet like our current styling users are and not all concentrated in one place or in a few countries so that's that's the other thing that comes with it so this is not an encouragement to go and say Hey you know forget about fiber I've had people asking me whether we should stop investing in submarine fiber cables and I basically went God no if you can uh you know keep on you know keep on going with this build your domestic your last mile fiber networks anything you can keep out of space is going to help get get somebody else connected thank you thank you so the the the last question I'm sorry we we have not time we have to be used next groupish from internet exchange Nepal and sanog um a quick question um so you mentioned that at any given time there are around 10 to 12 uh satellites that we can see uh on the space I remember an incident uh last year I believe that around because of the geomagnetics storm there were around 40 satellites that was damaged for uh starlink um in case of such damages in future uh what is the kind of a timeline that we we have uh so that the another satellite can be relocated to have that in position uh for a kind of a BCP situation okay so this is this is really a problem that shared across the globe um now the 40 satellites that got damaged I understand were actually satellites that were not on station but on their way up to station um and basically um styling is launching in the vicinity of about 40 to 80 satellites a week at the moment so um that is pretty much small fry for them and and also the satellites of course are distributed you're talking already at this point in time well over 4 000 satellites in space so 40 satellites getting lost is one percent of your service um so this is this is not a biggie and uh Starling doesn't look upon it as a biggie by the looks of it I mean they basically say hey we lost it but the other thing you'll remember is the way that styling builds satellites is to build satellites like other people build washing machines right you build a hundred or three of them don't work you don't worry about it it's not like building satellites in the past where your satellite wash the Apple off your eye where you would make sure that you had a super clean room that basically the thing must not fail well styling satellites is different they don't use clean rooms when they build them they just build a large number of them if a few of them get lost then so be it so a totally different philosophy there um Let me clarify my question again uh what I mean by uh that was at in any given time if 12 satellites are looking at my position I'm not concerned about 4 000 satellites but I'm concerned about those 12 satellites so what are the chances that those 12 Satellites May go down and that's how what's the kind of a BCP plan that takes uh statistics to relocate the new satellite in that so basically those those 12 satellites would be moving out of you within minutes and you'd be they'd be replaced by other satellites that are following down around yeah timeline that to take to relocate those things um probably in order to fill the the the the empty slots in the respective orbits probably in the order of about weeks um but then again it's not a problem if you're sort of having an empty slot in an orbit because there'll be neighboring uh neighboring orbital clients that can take over um so that so you it's not a loss that you would as an end user necessarily notice okay thank you does that make sense okay thank you I'm ganga from glb Japan my question is the The Styling provides like point-to-point connectivity or satellite or does it always associated with an IP address or so basically what it provides at the moment is two types of service so they provide a service so the common service that they provide is really direct to end user and that comes with problems like for example that there's no obvious way to put say for example a Content delivery server that conserves bandwidth on the space segment um on the other hand they're also Opera they also offer for a bit more money significant bit more money they offer a sort of business unit or high performance unit that can talk to several satellites at the same time and that's intended to provide sort of point-to-point links with the idea that you know basically if you're you know say for example having a drilling station out in the desert or something like this that or a little village or something like this so this can pour your village internet um but um that's not a particularly mature technology at this point in time and they don't offer it everywhere and where they offer it it seems to be getting reasonably uh mere reviews from what I've seen okay yeah I would like to understand more like when I say point to point means the two points are on the ground for me in my case okay no basically the only the only thing that they offer us at this point from from all I know is basically styling infrastructure to user user to starting infrastructure so if you wanted to connect two points you'd have to go do that through the styling Grant Network typically okay thank you for a great presentation and the discussion yeah let's move it to the next session please approach to next next session is uh over what is the interrupt tokens unit continuous challenge over the 30 years thank you very much for introducing my about me and I'm tacostomy from National astronomical Observatory of Japan and I'm also the next team member generalist of the winter of Tokyo Channel today I want to introducing you about the interview what is enter of Tokyo sonnet and uh project is and also so we have to are we we I want to think about what is the interoperability on the uh on the on the internet Technologies at first the the word keyword of the interoperability interoperability is the key for the extent severity or for of the internet so because of the the internet is made on the very uh highly uh distributed and also the anyone can connect the evenness this kind of device can use uh can running under the internet protocols so but the usual uh things will happen in under this and that equipment so so many equipment can develop by the server vendors but this kind of equipment can have need to be communicated with the other equipment using the several protocols under running and under the internet but uh so so we have to think about the interoperability within each other so the the for for this kind of Technology about the protocols is a really important to do the testing the interoperability within each other so uh this kind of inter operated texting is testing the running code and also uh on the status standards the standardization process and the ietf or the ITT on something like that they need to be uh keep interoperable within the several equipment of courses of from the several vendors so we this kind of testing process is really helping this kind of the study standardization process I think and also this interoperability testing can be helpful within the it well known each other or over among the several companies or several vendors because of the the we have to and will communicate within each other to solve the problem and if they want to make the inter interoperable to be the each equipment I think and the next is the I will introducing about the uh interface Tokyo the interv of Tokyo is the largest exhibition of Internet Technologies in Japan foreign this year the over 200 buses it's there and also the ones 120 000 people which that's coming to this exhibition this interview of Tokyo is a held in Makari message achiever uh in every year since 1994 it's just the one year after the Epic was established so this year end of Tokyo was the 30th anniversary so that we are just keeping this kind of uh of course this kind of exhibition everybody had in over the 30 years and also we have some we have interoperative testing under this exhibition for 30 years this interrupt Tokyo is the derived from tcpip Ventus box of held in Monterey in 1990 1986 and 1994 the this interview event is coming to Japan and this show net is a project of the very large scale demonstration Network inside the enter of Tokyo this annual activity demonstrating the latest Internet Technologies this sonnet is just a network but uh we are operating the one as inside the macari method the venue of the interview of Tokyo and connect to the internet and also that we are serving the connectivity to the to the exhibitors of the interrupt interview of Tokyo and also the visitors of there so this demonstration network is just not not just to demonstrate so we have to keep serving to the uh keep serving the internet connectivity to the to the exhibitors on also to the visitors uh inside the Show net we are conducting various demonstrations and also have some interoperability testing inside the shownet the this year the several keyword is there like the srv6 svpn mvpn best age the 400gb Verizon store plus and local 5G and more there are so many uh of course this event this interview of Tokyo is the exhibition of the internet Technologies so there are so many vendors uh have a boost inside the interest token and also this kind of this these exhibitors serving as The Cutting Edge new and brand new equipment or the products for the inter studio and we are gathering this equipment and build uh build this drive demonstration network with cutting this Technologies uh this topology map is the uh described about the network diagram of this year's shown it there's so many objects is there the this at this year the over 2000 uh equipment and services uh was served from the exhibitors that we called them as a contributor uh the these contributors as I offered to use this equipment and also can help with the helped by the engineers from their from their companies and this topology map shows the the upper side is the external network and the bottom side is uh very close to the network users and these red objects describe the layer 3 equipment and also the green one is the layer 2 and the orange or the year one is the security devices so the we are gathering this there's so many equipments to just do inside the Macarena the venue of interface Tokyo and we just we're uh building this network within the two weeks before starting the exhibition now after that after the exhibition finished we just clean up the venue and also we are sending back this equipment to their companies so the the this inside the shown that there are so many uh world and Japan's past products is there because of the interface Tokyo is a uh the exhibition of internet technology and also the commercial uh uh trade show so the so many companies try to promote the the brand new products at the interior of Tokyo and also we can running under the sonnet so the these steps can seeing the uh this brand new products can be running under the uh this derived network with the interoperability so the at this for this year uh inside the Show net we have three large fast products in there and also several Japan fast products is there so this is the uh very unique point of sonnet I think the sonnet and also the interview of Tokyo is the uh I was heading over the 30 years so we have the several challenges under the sonnet and the 1990s the inside the Zone had to be running the and also the testing in the interoperability with about the ATM and Metro ethernet is after running and the test after testing inside the sonnet there's so many several services or products is there and the 2000s running and testing the interoperability about the adsr and fdth and cgn and also secretness security operations center this kind of things is already running inside the uh the amount wall but this at that time not so many uh the several Technologies is not running in commercially when we are trying to use inside the sonnet but after that that we have some uh knowledge we have some marriage under this kind of uh activities from uh write this Cutting Edge Technologies so this the of course inside Japan there are so many Engineers from the telecon location company also coming from the vendors and some from the academic so the we get some experience to do the what how can we are running the how can we have working uh correctly this kind of new technologies so the we over the 30 years the sonnet project was uh continuous challenges about this kind of new technologies so the why is the unique point of sonnet the so this a kind of interoperability testing will be have maybe happened in several uh areas right the uh the severe vendor gathering in the one press and the testing the interoperability of the uh of the new technology or the new protocols but this kind of large scale demonstration Network can be uh it can be made so many vendors or so many Engineers is there so we can discussing the how can how should we do under these new protocols or the how should we do to solve this problem or the improving this new technology or the protocol so this kind of discussion is really important and not only the discussion just the running equipment is there so we are just testing and testing testing to do to this new technology should be worked properly so the Show net is a really good place to testing for uh testing on new products with multiple vendors because of the the usual network is made by this not so many vendors but this uh sonnet is just gathered so many products from the server vendors so that so it's a very unusual Network I think but it's a very good press to have a interoperability testing with with the several equipment and also with the server vendors and also it's a very good point of the show head is it's running in front of the uh visitors attendees of uh about afro of the interval Tokyo so the visitors of interview Tokyo can sing the this product is right and correctly running and also this kind of new technologies or new protocols can be running under this kind of this mix a mixed environment so the I just some of the some of the key topics uh I will introduce you about this year uh this year we have several uh experiments about the uh run the connection within each other the this is this right showed about the real integration by IPO of a dwdm dwdm is a usually this current architecture line shows about that uh the the equipment connects to the transponder and the transporter changing the web ranks to the to to the optical network dwdm network and also and Rodan gathered these uh Ramada pass and connect to the each other but this year the we have IP over the DM so this small coherent Optics can be installed to the network devices so network devices directly connects to the dwdm network this is the IPO but the difference of Ip over data dwdm this device is a high power version of the 400 gig the the data plus transceiver it's a brighter version of the normal uh 400 gigs data so this here uh this the this kind of equipment is uh going to be the drill is small and we can install to directly installed to the network equipment this year this uh 400 gigabyte data plus device installed to the Cisco router inside the sonnet and this Cisco router directly connects to the uh DW DM Network right the ion APN uh it's the ion PN is from The Entity and the directly goes to the auto match via ion APN Network and also this equipment can be installed directly installed to the network devices so Network device can measure the parameter of the wdm parameter of the wfdm so the usually this the optical uh the the Lambda networks can be handled only the ram the network equipment but we can now but now we can handle this kind of things by the network devices investigation is a very useful for the server Network I think but we are testing the several types of the Adidas mitigation inside the Sony and this mitigation can support the 100 Gig rings and the uh this uh these devices running under the 100 Gig rings and APK is a really frequent to be discussed in the Epi net meeting I think and we and on inside the Show net we are just uh we also running the rpki origin validation in from the uh inside the Show net about for the uh throughout of the internet and maybe the last time uh that that was the 2016 or something like that the sonnet also trying to using the rpk validation but uh also this year uh this is a very slightly in a studio increasing that so that the that can be buried validate the uh by the validate that route from about the internet throughout and also we are challenging about the free VPN best backbone the the recently there are so many times the where we mentioned about the segment routing I think and also that we are running the srv6 inside the sonnet and also access net uh also for the access network we are using AVP and breakdown on topics of the S7 VPN and so we create the vrs mean PE and connected between vrf uh by the srv6 and dt4 and dt6 so this uh VPN based network is very unique because of the 0.2 Point Rings inside the backbone only using the IPv6 ring clock address so of course the ipv support address is already exhausted but the IPv6 address handling is a little bit a little bit to use that because of the uh the address is really wrong I think so but we this kind of the VPN invest backward only uses the IPv6 wrinkle rink rocker address on the point to point ring so we we are now we are free from the address management inside the platform so they're inside the equipment and at the 0.2.3 interface that we just we only just this uh configuration under the interface and the srvcs so we can handle the so many uh the the new technology right there for example is a frequently discussed and write the networks Rising so the real uh you we are using the flex Argo by the several metrics and we can handle it by the routing process the routing protocols and also have uh interoperability testing about the srv6 microseat so the the the usual srv6 needs to be directed The segment list inside the IPv6 next header but now this technology like microsit is supposed to encode 2D IPv6 address of about uh about the series so that we uh sometimes the the extension factor of ip36 uh not can be handled by the several network devices because of the this is an unusual behavior so we can now encode to the to the Citrus inside the destination IPv6 address and yeah yeah and also this uh is just discussing inside the ITF so the the several uh implementation have coming from the severe vendors so but this implementation must be have uh interpreted within each other so but sometimes not so we have to checking the interview TVs in each other and also can find what is the problem about is the current problem under these Technologies and relative extension for the access network using the EVP investment the evpn version is already been on about this Technologies but we are just conducted this EpiPen application network with the six models from three companies the last station the situation said there is a really hard to make the interconnection using the evpn within the several products and also the server vendors but the of course there's a real problem is there so the but we fix we fixed them before the exhibition and we are running this a EVP and vxron network inside the sauna okay so the sonnet is several challenges is there in every every year the at this time we we I only have the limited time so we are just introducing you about the radio 2 and radius technology challenges inside the Show net but we have the several categories inside the Sony the in the every categories have the every child every challenges for The Cutting new technologies and so this challenges can be made the uh several interpretive testing within each other so we have some we have get some knowledge is from this uh type of experiment and so there is a really important things for the internet Technologies so we we uh we want to keep this kind of challenges for uh for a long time I think that was my presentation thank you very much thank you very much maybe I we can accept one or two questions from audience so okay then there are let me go to the next session okay thank you for our presentation please address you to the two minutes time again yeah last succession is uh Jeff Houston from Happy nick uh the title is a good party shipping afternoon all um God it's late in the day I don't know about you I'm struggling to stay awake um let's see if this can do something about it all of you know what TCP is you are all struggling to stay awake aren't you TCP God knows um there's something odd going on on the internet today and for any of you who run Networks you may have noticed that we're getting awfully good at encrypting everything um these days around 90 of all internet traffic out there on the public internet uses encrypted payloads https and the equivalent you know TLS is ubiquitous um 70 of all traffic comes from cloud servers near you so if you look at the amount of packet miles to deliver content to users most of the traffic is incredibly local everything you're looking at right now came from a data center in Osaka just down the road there is no routing there is no Transit there's no such thing as routing security if you're not routing most of the traffic we're now using is incredibly local we actually don't use the network we don't route and routing security is irrelevant because we've managed to crowd all of that data all of that service under your nose and wherever you go on this planet you will find data servers very very close to you except in more isolated areas which are either poorer or incredibly isolated and maybe starlink will change all that too so this is a rapidly changing Network that's getting smaller and faster at the same time and these days right now around 20 of all traffic users quick now these statistics all come from Cisco there's a reference at the bottom of the slide if you don't get up later but this is actually based on observations done on real networks what have we done to TCP that we're now walking away from it moving into an entirely different world of encryption and using a different protocol so let's go back a bit why am I saying that TCP is dead when an actual fact the internet wasn't really just a datagram protocol way way back in the 70s when we were playing with these kinds of things there were a whole bunch of different ways of packetizing networks the itu took up one with the victims called the ccitt with x.25 there are a whole bunch of these

things that just sort of chomped up data into little segments and then sent them on their merry way why was the internet so special I actually believe it had nothing to do with IP IP was irrelevant the true Workhorse was actually the transmission control protocol and what the transmission control protocol did was actually half Technical and half economic because the whole problem in sending packets is that if you've got a chunk of data Big Data and you want to reproduce it you have to make sure that every chunker every packet I send through is reliably received you need a reliable streaming protocol to make all this fly and that was the job of TCP okay TCP was brilliant it did effectively congestion control it sort of went as fast as it could and no faster it fixed up all the errors in packet loss it lifted large buildings but the beauty of it was the network service provider didn't have to run it you ran it in your end systems all of the cost of building reliable Communications was outsourced away from the network and pushed off into your pocket as users building internets was cheaper than building frame relay or ATM or any of those other ones that tried to do reliability inside the network and the beauty of it was one size fits all because this thing's self-clocked it wasn't 10 megabits Ethernet or 100 megabits or one gigabit or anything like that it just ran as fast as you could shovel data through the network incredibly adaptive and we've done a lot with it over the last 40 odd years it's incredibly efficient if you tune your end system whatever capacity is in the network will get sucked up by TCP it just does it brilliantly it's even slightly friendly that if two ccps are trying to compete with each other on the whole they're largely equilibriate TCP is not a network it's a client Behavior an actual fact when I look at servers and clients the client at the other end has nothing to do with it it's the server that does everything it's the server that figures out which congestion control algorithm you're going to use and how you're going to use it and there are a whole bunch of congestion control algorithms out there and effectively it's the server's Choice your device can talk all of them because it's not the determinant it's the receiver so different TCP behaviors do different outcomes if you feel like stealing all the bandwidth around on a particular day run bbr it's really good at that if you want to be slightly more friendly to everyone else and you know run cubic Etc so it's incredibly flexible and we've been pushing and prodding at it for many many decades anyone remember ttcp some of you should um you know how you do this freeway handshake and then you send some data and then you do a fin what if you put data in the opening handshake hi this is me and here's some data let's complete the handshake and run away it was a really really good idea I seem to remember Bob Braden was very very keen on it totally insecure but a really really good idea the insecurity killed it multi-part TCP where you notice everyone now has things that have Wi-Fi and cellular and everything else on the same device why don't we use all of those interfaces at once and oddly enough Siri um because someone wasn't performing adult supervision at the time actually does do uh multi-part TCP and if you look very closely and you are connected to both Wi-Fi and cellular Siri will actually run over both channels at once really clever window inflation case performance explicit congestion control signaling it's all there and you go well what else did you possibly need this stuff you know slices bread and toasts at all at the same time but there were things that TCP just can't do um applications can't run their own TCP because on almost every platform we know the kernel of the operating system whether it's Unix or whatever Windows is these days grabs those packets and says mine and users actually find it incredibly hard to write an alternative TCP implementation running up in userland it isn't multi-stream and these days with complex content for example if I'm getting a video stream and it's broken into chunks and all the chunks are coming from one point I can't drive them in the one sort of end-to-end encryption state but still treat each stream separately it's not multi-stream TCP does head of line blocking um all of the control parameters are out there in the open and there's an endless parade of nonsensically insane um TCP accelerators TCP decelerators TCP mucking around things um rocket and a whole bunch of these other products at the time which were meant to make your TCP go faster slower or sideways or whatever it was playing around with congestion control parameters was you know a spectator sport for quite some time and TCP had abounded I remember some service providers trying to identify everyone running BitTorrent and making them run like molasses gee thanks end-to-end encryption although everyone uses it in TCP as an afterthought it takes another round trip time it's slower to start up it doesn't do reliable datagrams doesn't do remote procedure calls it actually relies on the application to do data framing it has no implicit framing it's a protocol of the 1980s and guess what it's not the 1980s anymore we want more these machines are more capable but TCP isn't delivering now what do you do you can keep on playing at the edges and tweak and tweak and a number of folk do and a whole bunch of PhD kids are sitting there going what if I did this to TCP and you probably could but realistically the core of the functionality isn't changing or you could do what's going on in the larger internet lift things up into the application world and just ignore the lower layers ignore the network ignore the platform ignore the kernel ignore TCP and build your own and that's what quick is build your own what you actually do is start with a basic UDP transport down into the operating system and the platform and down on the network so all this is just UDP nothing special everything else is running in userland in your application all that end-to-end encryption all the transport service whatever TCP was doing is now you're doing in the application data Integrity session control congestion control you can read this as fast as I can now some platforms like apple actually give the application an API going look we've done it for you you don't have to use it if you're Facebook you'll do oh no no no sorry my Facebook application runs my version of Transport control I don't trust Apple why not I don't want them looking over my shoulder I want to do the encryption I want control it's my data that I've learned about the user I'm not sharing with anybody ever and so the thrust to sort of go there has actually got sound commercial reasons uh there's a block diagram it takes some of the components that we do today you know TCP TLS and then the data stream and just simply smooshes it together so inside the one stream is quick multi-stream encrypted rpcs data Integrity congestion control all this wrapped up the underlying thing is simple UDP packets and on the top deep inside encryption deep deep inside encryption is HTTP you're never going to see it so it's so much more than just encrypted TCP over UDP so much more than that I can multi-stream like crazy I haven't got any head of line blocking it's faster by one round trip time so if you're actually doing a whole bunch of really short sessions rather than taking two and a half round trips to actually bring up an HT or TLS session quick just does it in one tlsr handshake and actually does TLS 1.3 the first packet can be data if that's the second time you're seeing it individual flow control I sound like a Salesman don't I it's pretty cool it is really amazing it does slice bread it uses UDP and in V4 these days all of you run um Nat so on on on your V4 networks and gnats are not UDP friendly they'll certainly do a binding but they won't keep it for very long why not because there's no fin packet there's no session for that nap to grab onto going I'll keep this binding for as long as you're active and so what most most gnats do is well open up a binding session for your UDP data stream and I'll keep it open for a while and then I'll shut it down but you're still active and the next packet you send is going to come out the other end of the NAT with a different Source address quick is address agile so I'm busy talking to you your 10.0.0.1 we're having a final conversation then all of a sudden the NAT changes state and your 1003 how do I know it's still you inside the encryption envelope that you and I are sharing as a secret we use a persistent connection ID and if the packet with the new source addressee comes with the same ID I'll start to think I think what's going on I'll send you a challenge if you can answer that challenge you're now 1003 and we'll just press on so this stuff works across Nats seamlessly no stun no turn no ice none of that crap you just simply work across Nats in a stateless fashion it was what we should have done 20 years ago so it doesn't fragment why not because if you run V6 and try and fragment give up all hope is lost fragmentation and V6 is the work of the devil it's never going to work it never retransmits packets which is really quite odd but we alway

2023-09-19

Show video