UKNOF40 - Scaling for Ultrafast, G.FAST, FTTP, 5G and the Cloud

they slum near McCrea look after em PT's technology strategy the kind of bizarre title of chief architect which is I still don't fully appreciate what why or whatever what that's for binary and I thought I'd start with a picture of the sr-71 who's heard of the sr-71 the fastest the fastest Arab the fastest plane where you can breathe air on it um and what's interesting about the sr-71 emulation to the talk I'm a bit of a space not as many of you may know you should definitely try to win that Apollo rocket is the sr-71 was all about you build an aircraft and actually was built in secret even secret from the US government and actually was made of titanium the u.s. than have any titanium so most of the titanium and the sr-71 stolen from Russia which I thought was incredible given it was the 60s and the Cold War was old but it wasn't just about building the plane as a build it was about building a new way to build and that's what that's kind of the the the thinking I tried to give to to people who are building networks we I think in our industry we kind of we have too many people saying we've tried that we did it that way we tried it before and actually often if you try again or just change it ever so slightly and things can be different and there's a boot called sled driver I think it's cold and by one of the pilots of this I really recommend reading this phenomenal book um so back to networks and this clicker isn't working thank you and next slide please and so I thought I'd just start a little bit with them they are indeed that we did BT that I am I'm kind of the lead sponsor for in terms of the output and we invest a huge amount of money in research and development we and and if you go onto the next slide and that's actually pinned by we pin that research down to what we perceive as key areas of improvement in the network um and this just uh it's just that the ones I could remember was creating the slide but if if I think about if I think about we talked about telemetry and and we're using that for autonomous healing in the network actually this year are the voice net won't believe it or not hits at five nines for the first time in awhile and and it's a technology that's almost as old as their so our 71 but it's a it's a platform that it's got huge role I you know pretty much every traffic light depends on the voice network and every missiles carrier nuclear power station depends on the voice network so it's important that it's reliable and there's things that you can learn from that from how you run and operate the voice network and I look at the voice network is a is almost like a cloud it has a whole bunch of processing units and a control plane and and it's distributed when you have a problem with it few people know about it and it'd be great if we could get that capability more into the network and we can analyze that so these are some of the partnerships of ours at Cambridge University yesterday anyone heard of graphene yeah and graphene and telecoms and so graffiti and telecoms is very interesting because it's a you can use it as I guess the words like an accelerant or an agitator on silicon which means all of a sudden the optics that we can only drive a certain space speed suddenly can go faster if I give an understanding and our network we have some segments on our optical core where we've used 96 wavelengths and we have to use another fiber again another 20 or 30 can be really valuable we're doing a lot of 5g research with Bristow that were building a little 5g network in Bristol and here in London with Kings and I think the point is is experimentation and research is something our industry needs to do a lot more of not and I'm not talking about the vendors I'm talking about ours in partnership with universities in partnership with with vendors and open source and other organizations and our network just quickly we have just under 2,000 Ethernet nodes and we have super super fast fiber accessed I think 92% of the country our core networks 106 sites we have MPLS to 1800 sites we have a high capacity optical platform that's on 106 sites and is extended in London for the Metro collective a we typically operate a three exit Network and which in some cases and always three exit because of the fiber topology there are some parts of the world we are doing a third route requires you to make a mountain go away so that's that's expensive and and and what we've been doing a lot more of is and I showed to the Comcast guys because I think what they've had to cope with firms of bandwidth growth over the last few years is pretty unbelievable and some of the things some of the things they didn't do that I'm thinking we should be doing and about multi-layer modeling and models in general are really important and we also carry a huge majority of the mobile traffic in the UK as well and obviously we on EE and this a bit let's say this is kind of how we we think about our network is very simple at some some it's very simple but allows us to talk about it and actually language is an important thing you hear quite a lot of language today where people mean something slightly different but it starts off in access so we have open reach and we have radio access and through wholesale we have EDS l that goes into MSC which is a Nokia 7750 and then we have our P network which is as is 212 routers of 106 locations and with 100 Giga factors one side it's got 400 gig tests on it right now and and we run you know bog started MPLS wide pvp errands for some different services so the mobile network is on a VPN for example and we run Ethernet across many of your customers of us and we appreciate your business and your challenges because our goal our goal is to make networking better and make the world achieved what it wants to do through networking that's why I work at BT and you can see some of the paths or um and I've spoke about this before if you look at one of my previous presentations I showed our population and traffic a population coverage of the UK you know everyone says why is all the traffic and focused on a southeast so where all the people out of course and it really is that simple begin to what we are in what we're starting to look at is this does IOT or machine-to-machine um technology driver differently we also have a number of regional hubs we sure they are and where we see some element of regional traffic but it's it's it's in the margins um so I talked a little bit about there's often a lot of talk about what we should be building in the network and and actually um you know clearly that where we go with the network and I've worked for some companies clear where we go with a network has to be profitable for us as an industry not just for BT and and I've seen I've worked for many companies as I'm sure some of my colleagues have well you know we spend a lot of money build a network and never got the return and what happens very quickly is is it's a race to the bottom and you go our business or you're required so I worked for Cable and Wireless we couldn't invest we can make enough money from the network and and we were they were acquired by voter fraud because there was just nowhere to go what this shows is the kind of growth that we see year-on-year so the top graph is is how we've seen traffic ignore the bottom bet the bombuh is just a year's ignore the day on it should just be able to April which one is sure a year and then we're trying to show that the traffic peak time fiber usage and per user so if you go back to 11 12 which is the bottom line you can see um what what you know the difference so that was the you know the UK Olympics which doesn't feel that long ago and and you can see that the traffic growth year after year it's pretty significant what we what we tend to see is from Kenneth octo ggest tiller by says Easter here but it's probably more like March early March we see a lot of girl fan and it's kind of quiet until they August again and actually for us that's very helpful because our financial year starts and April so we can build a lot of capacity in April and get ready for the kind of surge that starts usually when the schools go back after the summer holidays and you can also see in fiber in copper and yeah there'll be some smart guy see now is that fiber really fiber as a copper and it's it's effectively our NGA platform so what we see is you know a huge we still see a massive amount of growth in copper usage that's ADSL and the interesting thing about that is um you know the world in these armies out there saying everyone should get fiber but we have many customers and who could have fiber today a fiber from the cab that choose not to and it's a significant number perhaps don't economics perhaps down because what they've got walks walks for what they need today and I'll show you an interesting start on I saw my whole network and layer on a CDN so we've been working with the CDN since 2011 when I joined BT we had a bit of a mad policy which was we're gonna try and make money at the seedy ends I said no other thing that's a good idea whether we work with them and save a lot of money as you can see more of most of our traffic is CDM traffic peering you know we all run around and have lots of peering conferences it's becoming less of an issue less of a thing for me to worry about frankly I can kind of see that feeling away over time as we get very much is that is the the network itself evolves away from this kind of mode of working that we have today and and and places like you know even the links were you know I was on the board there for for 15 years that you know the value of that in our network is it's important but it's not what we're constantly thinking about Imperia interconnects not what we're constantly thinking about anymore it's much more about bringing quality and connectivity and to customer we also use multicast to deliver TV I think I've talked about that in another presentation and I thought cover I was the sky guys did a great presentation last time this is Manchester say and vs Liverpool I think there's if you're a Liverpool fan its provide great day a smile being in Manchester because the one bit was the biggest event on our network so just in broadband this doesn't include anything else and we were carrying 10.3 terabytes a second that was served from a few CDMS are on CDN and the guys at Akamai who do a fantastic job and and and and actually what that and actually that wasn't that that wasn't a high number of customers so concurrence is the thing that really keeps me awake at night if someone came out with an app and I cannot look it for enough innings played the game fortnight where you get huge concurrence like a zoop there's a an event a few days ago where there was just massive concurrence and people play in that game drive and it and luckily it doesn't use a huge amount of bandwidth but hiking currents high bandwidth services I've just lost the screen eyes back and hiking currents high high bandwidth services are very scary and fortunately not everybody likes football and otherwise you know it gets hard um the last thing that was big like this there was high concurrence and high usage really huge thing and was the guy who jumped out the red red bull Bullard luckily YouTube only had standard definition then and otherwise that would have been a very bad day and um this is an interesting graph I'll let you think about it for a minute well this graph tells us is and we have a more detailed version of it unfortunately I could I wasn't I wasn't I was kind of said I can use it but leave my past at the gate on the way out but basically what this graph tells us is that and we see more users even in the in the bottom end of the network kind of up to 24 megabits per second really starting to hit the line speed much more often than they did only a year ago and what this what this is telling us is that we have we're starting to see the ADSL type technology become you know a limiting factor to consumption of the Internet and back you know and actually I was taking quite a long time if you think about it if you think about dial-up it was almost it was almost game over from when we started dialing up and we could never get a fast enough more than and but what's interesting about it is is it's not really driving a change which I kind of I kind of struggle with which is oK you've got 20 Meg and you and you're hitting at 20 mega hit night when you make limit quite a lot and you can have five-hour why aren't you buying it and it when people talk about investing networks and build an effort FPTP and you look at that and you know it starts to be it starts to be really challenging basically but you can see that the curve has swung from October 2014 to September of 2017 and so this is actually this is M this is my line at home and I've got more bandwidth than probably most people including some of the companies at home and and I never this is this is that this is from March to April the last month and I have 40 megabits per gigabits per second to my house and and forty five Meg's my peak for the last for the last month and so you know when you look at all like that and I'm a heavy user I mean I've got all sorts of devices I've got an Xbox that seems to update every my bling Kathy actually last week my my TV decided they wanted to download an update and most of these most of these Peaks are some sort of patch or some sort of update and and you know it in an in here at a I'm a huge pinball tournament in my house and I gave everyone I wanted to really hammer the line I gave everyone you know the best access I could give them and and it's not even noticeable on here um which is which is kind of peculiar um anyway so the goal is you know so you kind of think maybe you know maybe some of the pricing so clearly we said we all in this industry for the most part sell products where if the customer consumes more with it doesn't necessarily translate to revenue so we were really focused and actually most of the work that were focused on as I'll are getting more bandwidth for the same price or getting more bandwidth for last price since 2012 we've taken 70% of the cost of our network out um and that's predominantly on our architecture which was quite complex and actually was a very focused on voice to move into an internet focused architecture and we're about to deploy the full generation of a factory we are deploying the full generation of writing engine 421 see we've got 200 gig waves in production we're about to do 400 gig waves into production we've got an optimized metal core in London and and we run our WDM platform that's that has one optical plane which has made life a lot easier for us and and some of you guys that buy from it um and it's all really been focused on unit cost pounds per Meg and everything that we've you know but probably 60% of the work that we do in network architecture is is focused on that and this is our next generation rotating platform as a Nokia SRS and it's a petabyte scale right on and we started putting them into the network and so far so good we currently have the the main workhorse of 21cn is 7950 XR s you can see and that roots are there is probably the busy it may be although Akamai or or Netflix might beat me on this but it today it will hit 16 terabytes a second and it's in central London it's a pretty busy box and me over saying on its in touch this and die [Music] or we will have and so it's you know trying to you know it goes back to my earlier point which is we've kind of we've built networks you know bigger box bigger box bigger box actually the bigger boxes are becoming a problem and let me do this slide first the bigger box becoming a problem because not only do you need bigger boxes but you need huge infrastructure we managed to close a bunch of Main Roads well so we installed new chillers and to a glorious place called being our of house Nigel was probably having a heart attack right now just thinking about it and we are you know the cooling requirements and the power requirements for these devices is just insane I mean what per gigabit and kilowatt per rack et cetera is you know the they're all kind of going in the right direction but but it but it's basically more of more of more and actually you know what and then and then you've got to think about our 5,000 exchanges with with roughly five at least 5,000 ups he's if not 5,000 battery sets and probably 10,000 January hours they all need to be maintained the off to be upgraded minutes in it's a it's a it's a big effort for us and it's expensive and and again so one of our things that's big kind of journey for others and over over the next probably 10 years is exchanged closure and we've done three or four of them to kind of test over the last few years and it's closure or moving on and you know we know what the opportunities are we did a model actually working with Bell Labs and if you were to build BT today how many locations would you be in you could probably build it on 800 locations as opposed to the the five and a half thousand that were in today and was talking about about 5g Asheville it was a slight a mess we're also gross of looking at and was created by this the optical vendors are finally realized that building these big boxes or is madness and building kind of slim you know very focused devices used for datacenter interconnect you can see the bomb as a four hundred gig transponder that we've been testing and seems to work and and the but began more and more you know getting these devices smaller unfortunately increases the heat so it's kind of a bit of a trade-off but you know we're working with a bunch of suppliers in that space and also in white box which are come on to a minute and so 5g we just acquire some spectrum and we've we've got we're building out right now in London and and also in the same time rolling out much more into 4G with more carrier aggregation in London you can if if you've got the right phone and a certain London they kind of baseball cap area you can get 800 megabits per second on your phone and again that you know in London we're already busy and we have to cope with that it puts a lot more effort on to the the core and then we'll do massive MIMO testing which is really important for higher frequency spectrum because we need to bend it around buildings and our core network for 5g you know we don't talk about fixed or mobile and where we talk about the core network and I think increasingly that's going to be important and it's why I can I flagged the thing about you know don't assume that you've got diversity and wireless and non-wireless because you probably don't and so a little bit about network loath I'll skip this slide and this one jump to this one and so you know Rob Shakir who's used to work for us did a great presentation on segment writing we're pushing very aggressively because we need to scale to kind of need to scale to the dense you know imagine every traffic light instead of being connected to the the traditional transmission network is like linked to the IP network you know you think about how many devices you need to do how much control play and how much optimization you need and Optima I we believe that probably just optimizations probably worth 20 percent of our investment if we optimized it perfectly and using segment writing with MPLS ASR over UDP into the data center and we think will give us you know a lot of control over the traffic with path M path computational and devices that that will dynamically create paths between the network and anyone who's rolling out MPLS te please stop and do this because we've done MPLS T and it's rubbish and telemetry this this so this is a problem is that they kind of dated back to the Olympics um a actually came up with a solution we just didn't have time about it well the thing that kept me awake during the Olympics was 10 seconds of Usain Bolt in the hundred meters because we wouldn't even if think someone had gone bad we would know about it for five minutes later whilst SNMP had its party and so that's where the telemetry idea came from and it came from you know if you think about it and this is a crazy thing NASA were doing this in the 60s there are a whole lot of stuff coming to our mainframe and their mainframe this idea what was important and flagged it to the people why can't we do that with networks today and so we've been working with the on telemetry now for almost three years starting to coming to code and if you're not looking at this I really encourage you to look at it and because I think this you know everyone talks about NFV Sdn or bunch all or nonsense I believe that this is the most important thing that we can do in networks today and I think I'll have the biggest change and it's great if I do only and we got a benefit from that bad quality the industry to tag along and this is kind of a high level so we have some machine learning on our telemetry platform where we'll see something happen so solve it the machine learning we'll see what what what we did to solve it and it will do it again and we have this running and our operations center up in our actual park fairly early days there's a lot more to do and we're not you know machine learning any I is is uh you know there's a lot of ranting and raving about it but there's probably really only two companies that know what they're doing in this space you know we use know the the the usual open-source tools that have been mentioned before although we've decided to write our own deployment software because we didn't we didn't rape salt or ansible or any of the other ones that right there are because of some of the things that we need to talk to because I don't want this shift apply to the future I needed to apply to my legacy because probably the voice never was going to be here for another five maybe six maybe seven maybe eight years and white box switching so again we of this almost the deployment is an aggregation layer were working with bear fruit on this as well because I think white box which in just takes my reliance on vendors away and gives it to Broadcom or Intel or something like that so I want to have more flexibility in that but this stuff just works now a huge effort by the the the software guys to to drive the right quality it's incredible the coil that's that's moved on this and we're also working on this on optical platforms including head end technology for HTTP right and that is my talk thank you very much thank you very much Neil do we have questions I mean Neil has whistle stopped so many fascinating topics do we have any questions from the floor there's one over here in the middle of Trev very interesting new link so you threw in so much at us there it's difficult to choose a questionnaire but at one thing that caught my attention was the 800 mink megabits per second speed to it from a phone right yeah what I mean what applications do people envisage that kind of speeds from film especially when you think that you're only doing 45 makes at home snapchat of course have you not seen in the city there's like there like storm troopers the hundreds of them and they're all snapchat at the same time and you know it's a really good question mastic and as I prepare the roadmap and planner or plow record and all the capex you have to spend often ask myself the hundred Meg's what we're gonna do with that and you know but I think I'm kind of a believer if we had that same mindset in the early nineties you know we wouldn't have an internet you know we built a lot of this not not knowing what it would lead to so my hope is is that there is a use case for it now it's not that we build build you know there's an element of building it will come but actually it's also an element of demand and we will so yeah and gay and Romero and your phone I'll probably some probably move around a little bit as we get more customers connected with faster fortin's I mean the big but the big challenge in phones is battery life of course so that does provide a limitation bar you know perhaps we can use some of this technology and hard-to-reach places and I'm not just talking about rural there's many hearts of each places and urban locations as well we might have time for just one more one more question gone in the middle all right Neil in your you mentioned the stats where you've got people who are still on ADSL even though they could get V DSL connections and fiber some of you FTTP okay so do you have ignoring the FTTP for a second because that could be availability cost etc do you have any stats on where because of line distance lengths if they were to upgrade it would either be about the same or worse is there an element of because I mean so you know assuming assuming the koppers of the quality should be in a northern places that is and we're you know we tried to fix that as quickly and as best as we can but assuming that's the case there's very few places as tiny there might places that wouldn't see a benefit you know it's just customers not taking it out yeah and and you look at you know I mean I have a look at this two other guys that play in this space guy and talk talk well you know a big chunk of their customers are still on their DSL platforms as well there is a tranche of customers that don't live the world we live you know that Netflix if it plays a you know a slightly blurry rate actually is okay for them and versus you know many of us live in a fortunate world and some people don't and and that's why we reducing the cost constantly is really important but also I think we need to think about the right way of of bringing the cost I'm so we so that we enable it for more people and we're more inclusive but at the same time not doing that before we can actually deliver it and I think there's some actions that we've taken in the past mostly through regulation where or will make this really cheap in market a or market to or market wherever is all you've done is make it harder for me to fix it and and it's you know I can't stress how important is is that if you're not able to generate cash to invest you've run out of Road in this business incredibly quickly and and you know dies is the big Network challenge over the next ten years and I think being opt-in and getting the most out of it I think we're a lazier organized and relays the industry that doesn't optimize in the same way you look at how BMW meet cars enough study that they have they have a bunch of things watching how cars are made and you know in every two or three weeks they'll say okay rather than putting the tire of my verse put it on like that we did none of that stuff predominantly you know some some some some places do it like some of the cloud guys but in the network's basis it's it's it's not as we're not invest in that as much as we should do found on that note I think we'll have to leave it there thank you very very much Neal [Applause]

Channel: UKNOFconf

Views: 1,750

Rating: 4.5789475 out of 5

Keywords: UKNOF, UKNOF40, ISP, Network, bt

Id: jgb_1axXuUA

Length: 30min 48sec (1848 seconds)

Published: Tue May 01 2018