These are unedited transcripts and may contain errors.
The Opening Plenary commenced on Monday, 13th May, 2013, at 2p.m.:
ROB BLOKZIJL: If everybody is kind a seat, then it's time to start, I think. Good afternoon, welcome to Dublin. My name is Rob Blokzijl, I am the chairman of RIPE. It's great pleasure that I welcome you all at our 66th RIPE meeting. I see some faces which I think might have been at the 27th RIPE meeting in Dublin, but that is a while ago. That time, we had 170 people attending, which was an absolute record, and this time, we might be breaking all records because this morning we had 570 people registered. But we will see by the end of Friday how many people actually attended, but I think we can already say it's a well?attended meeting.
RIPE meeting. RIPE is about operational coordination of the Internet. Sometimes we forget about that when we dig ourselves deep down into minute details of policy proposals, but I think we should never forget that why we are doing all this; we are doing this to further the development and operations of the Internet, and we are not a Law Society trying to come up with the perfect legal system or something like that. We are here as mainly operators and other people interested in operational coordination of the Internet. So, in the week ahead, I think there will be several times where we should remind ourselves that that is the greater goal and not whether your word makes it into the 7th version of draft proposal of this or that or the other.
Keep the big eye open.
As you will see from your programme, we have, again, a week full of meetings, roughly divided into two parts; the first two days are dedicated to a plenary session with presentations of a more general interest, and these presentations do not fall out of the sky by itself; we have a programme committee chaired by Filiz, who will, in a few moments, say a few words about the working ?? the good work of the Programme Committee. The second half of the week is dedicated to sessions of the various Working Groups. As you can see on your programme, not everybody can attend all sessions because there are parallel streams, so pick the thing which you think is at the moment the most important for you or the most interesting for you.
Last but not least, before we go into the real work, I would like to introduce and invite John Boland, CEO of HEAnet, that is the right way to pronounce it, all other ways are wrong, because it stands for higher education and academic network, which is a nice name because it goes back to the very early days of the Internet. The Internet grew out of mainly university and educational institutions where people started playing with this exciting new stuff called TCP IP, and their friends would say "no, no, no you should say Internet if you want to know more about that word, I think my good friend Denis Jennings, sitting here, is one of the inventors of this word but not only the word, also some of the underlying principles of the Internet and I am very happy to see him here today, welcome, Denis. Over a beer later in the evening you can find out from him personally all the ins and outs.
I would like to invite John Boland, who is an experienced opener of RIPE meetings because it's the second time he is doing this in Dublin, to come up to the stage and say a few words.
JOHN BOLAND: Thank you, Rob. Indeed, yeah, back in '97 I was the systems and networks guy in Dublin City University and the RIPE meeting was being held there, and mike Norris who I am sure a lot of you know, he is our ex CTO, he was organising it then and he would ask me, if it was at my campus come and say a few words and I remember saying could you show me the agenda and have a look at the programme, don't worry John you wouldn't really understand it, just say a few words. So six months later HEAnet was formed as incorporated company and I was appointed CEO of it so that is 15 years ago so hopefully I know a little bit more about it now but Brian Nisbet might argue I might know an awful lot either but it's a great privilege to welcome you all here, I understand we will have expected to have the biggest meeting ever in excess of 570 people, so that is great to see.
HEAnet has always been a supporter of RIPE and the RIPE NCC and the community Mike Norris, as I mentioned, Niall O'Reilly is here, Denis Jennings and Brian Nisbet and Dave Wilson are folks who have always been heavily involved. Over time it's changed, you have to look at where it was involved in the academic side, if you look at the number of global companies in Ireland now and the people who are getting involved now, so it's changing fast. And for us and our customers it's changing fast, too. One of the significant things, I suppose, in Ireland is that we now connect every single school in the country to our network, so we leverage our higher education and research backbone to connect every school, every student in the country. With that comes some challenges, and our Department of Education, want to us do this but they do not want it to be in a dangerous sense, so we have to provide content filtering to all the schools and that is a very challenging thing. I know that is the one thing that will get me on the front page of the evening papers if that goes badly wrong for six and seven?year?old school kids. In any case, we are providing that sort of connectivity. In the universities, things are rapidly evolving, too, and it used to be we had a sit down and try and plan the services that the students are going to need in the coming year but in talking to some of the biggest colleges now and we talk to them about what services they want from us and they say, well, actually, you know, we can't predict any more, we wait and see when the students come on campus with all their PDAs and we have to respond to what they want. So it's a very rapidly moving, quite a different environment. And then for us as an NREN as well, we are challenged to move from the networking or comfort zone up to delivering services so the resilience, it's no longer about the connectivity to the individual colleges now, it's all about international resilience, the virtual learning environment may be hosted in Amsterdam and critical 24/7 for students in the colleges, so it's a different set of challenges as well, that end?to?end service level agreement is critical. So that is where the importance of the work that you are doing. Rob talked about the future development of the Internet and so on so, for us with joined up networking in Ireland that is a very important thing. So, I suppose, two things to finish up on:
I am delighted there is so many people here. You have heard all about our financial problems here in Ireland so do be happy to spend lots of money while you are here, OK, number one. Help the economy.
And number two, just wish you all well with your work. It's important and HEAnet, and I think the NRENs need to stay involved and committed and we in HEAnet certainly are, so I wish you a successful conference, thank you.
(Applause)
ROB BLOKZIJL: Thank you, John, for your kind words. Next, we will have Serge Radovcic from the RIPE NCC, who will have some announcements and after him, he will introduced Filiz Yilmaz who is the Chair of the Programme Committee.
SERGE RADOVCIC: Thanks. Just while I am getting set up here, I have been ?? my name is Serge Radovcic, I am the chief communications officer with the RIPE NCC, I have been with the RIPE NCC for about 18 months now, I am one of the areas that falls under me is the organising of these meetings. And I am sure any of you that have been involved in meetings or organising meetings are well aware that when you are trying to organise a meeting that is in another country, there is one thing that is essential and that is to have a local host and I have to say this time around, as many other times with a host, it's been fantastically coordinated from your end, I think Brian has led a lot of that, so thanks John and thanks for all of the people you allowed us to steal time from to get this involved. So please, help me thank HEAnet for doing the work they have done so far.
(Applause)
SERGE RADOVCIC: OK. There is my little daughter. Basically I am here just to make a few announcements, the first one being that we are going to be running our survey this year, I am going to do another presentation in the Services Working Group, that is the second part of the Services Working Group on Wednesday afternoon where I am going to go a bit further into it but I wanted to announce the fact we are going to be opening it on Friday but more importantly, we have been doing a lot of preparation work with these survey in the form of focus groups and asked two independent consultants to help us, in terms of John Earls of New Leave consultant and Desire of the Oxford Internet institute. Could I ask them to stand up, I am hoping they are in the room. I think I have coordinated this well enough. Stay up for a second, guys. So you may want to have a chat with them during some of the breaks at the socials, they may approach you and have a bit of a chat about the work they have been doing and the importance of the survey and as I said, I am going to go a little bit further into it on Wednesday so you can sit down now.
Another thing I wanted to talk about was the staff at the meetings. We talked about the host, there is a lot of preparation that is involved in these meetings. You don't see much, you see these guys over on the side, there is a lot of administrative work, the communications department of, the web developers, I am probably missing a few people, but it's a really a company?wide effort to get these meetings happening, and keeping them running for week, so there is a lot of staff involved. We also bring a lot of our other staff from the service department that might not necessarily be directly involved in the setting up of the meetings but to talk to you guys from registration services, from customer services and so on. So, what we have done to help facilitate that a little bit is to, on the website, we have got a page here that you can go to, it's up on site, go to the RIPE and you can have a look at some of the staff so you may have dealt with these people when getting resource or chatted to them during customer services, now you'd like to thank them or kick them, either way they are here and you can have a look through that and see what their speciality S some of them are writers but they are doing a lot for v4 transfers and so on, that is how you will see some of the specialities. Please make use of that page.
Our staff are all wearing blue ?? can I ask all of our staff in the room to please stand up. These are a few of them. So, just stay standing up for a minute. They are not only wearing blue land yards, if there is specific information you need, please come to the registration desk, we have got the info had you been set up and people from registration services customer services go to our staff and ask them if you have got any questions. They are here for you guys so please make use of that.
What else have we got here? Yes, you can sit down now, guys. Sorry. If you are experiencing any problems with the wi?fi, don't just sit there grumbling and moaning and waiting until the end of the week, there is an Ops Room over here, if you are experiencing any problems please come and tell us as soon as you can. If it's too big a journey to go over there, let one of our staff know and we will send someone to you but we definitely want to get on top of any issues that we are having as soon as possible, so please, tell us now rather than waiting until the end of the week if you are experiencing any issues and by the looks of your faces, I am assuming that most of you aren't experiencing any issues at the moment.
OK. I want to talk a little bit more about the website. I am logged in on the RIPE 66 website with my NCC, RIPE NCC access account, and that allows me to do a few things. It allows me to go to the programme, look at the plenary and what you will see here, if you are logged in, it's going to give you the ability to rate the presentations, I am not going talk to talk too much about, the Filiz wants to talk to you the importance of asking you, to rate the presentations, but if you don't have an access account follow the instructions, it's very easy to do. Anyone can do it, you don't have to be a member of the RIPE NCC, you just need a valid e?mail address and you can get started. So it will allow you to rate the presentations, it will also allow you to go to the attendee list. And what you can do if you are logged in, is list ?? ask to view the profile ?? the photos of the attendees that are here, so if there is a name of someone you want to get in contact with you, I want to have a chat with Joe Bloggs, click on his photo and you will see. So you will see quite a number of the attendees have made use of this. You will only be able to see the photos if you are logged in, and upload the photos if you are logged in. Some of them have added e?mail addresses. This is pretty much for you guys to network with each other, we haven't designed it to be an open call support spamming of these people. It is designed for getting in contact with fellow attendees.
I think that is about it for what I was going to talk about. If you are having any troubles getting an access account or logging in or understanding anything I have just told you, please come and see me or any of the staff, I am sure they can help you out with that. Rob wanted me to make a couple of more announcements about the agenda. One rather large change that has occurred this week is the fact that the address policy won't be held over two sessions, they will ?? they won't need the second session on Wednesday morning, which starts at 11:00, they have been so kind to give that 90 minutes of air time to the NCC Services Working Group. There will be policy talk earlier in the morning, and in the afternoon there will be more presentations from the NCC about services, so that probably won't be reflected on the badges that you have hanging around your neck because we had to print these I think a day earlier than when that decision was made, but so a little bit unfortunate.
The other one that might surprise you or some of you haven't heard of yet is that there is a BoF, could potentially turn into a Working Group, the open source/Working Group which will be taking place on Thursday afternoon, so some of you may be interested to go along for that potentially historically making event.
We have got a couple of social events, we know you guys are here for the presentations but it's also nice when we get off the stage that you can mingle, so we have provided ample opportunity, also a couple of evening session where you can leave your laptops behind and have a chat with each other. The first one is in the hotel so if some of are you still trying to navigate your way around, that is going to start at 7:00 this evening, there will be enough signposts and you will probably hear all the noise so just follow the noise and there will be enough beer for you there and canapes, and something to drink. Tuesday night we have got a sponsored social event by ?? I can't believe the Irish Neutral Internet Exchange ?? in Dublin, of course. They have been helping us out as well with setting up part of the meeting, Nick Hilliard has been great and I am sure he has used a few other people ?? he is about to run out the door in case the network crashes ?? they are actually going to be sponsoring the event at 8:00 tomorrow night, in Dawson Street, it's not too far away, we won't be providing organised transport so registration desk and our staff will be able to help you on giving advice on how you can best get there. And Wednesday, this time around we are not going to have a social event. We heard some need back you felt, I haven't got any time to have a small meeting with any friends of mine, I always feel obliged to go to the socials you guys have organised so we are not going to have one on Wednesday, so you can either just have a rest or go have a quiet meal or walk around the city, I am sure there is lots of other things you might be able to get up to.
I think that is about it. I don't know if any staff are waving their hands at me or anything I have forgotten. It doesn't look like it, so Rob to save you getting out of your chair, I can probably introduce Filiz. Filiz took over as the Chair of the RIPE plenary Programme Committee six months ago, and she has been doing an absolutely fantastic job, sorry to already start on it, but I am on the mailing list so I see a lot of the work that goes on, and I see how many e?mails Filiz has been answering and I see how many times she has been cracking that whip of hers so she has been doing a fantastic job and I think you can see the agenda Filiz and the team has put together. Please help me welcome Filiz to the stage.
FILIZ YILMAZ: Thanks, and a big welcome from the RIPE Programme Committee, too. Thank you for the kind words. We do it as a group, this is a fantastic group, we have been getting new blood in and I think it shows, and we will have hopefully more about this, I will talk towards the end of the presentation, which is not going to take too much time. We want to give you a glimpse of what we do, who we are and how you can help us to make these even better.
When you look at the programme ?? the RIPE programme, overall programme, you see Monday and Tuesday sessions and also a little bit of the Friday as closing plenary, we have tutorials, plenary sessions and BoFs. Those are basically the legislation, if you like me to call it that way, of the Programme Committee. So the content of them are evaluated by us, received, submitted by the RIPE committee, obviously, the individuals of the community, and and then we evaluate and select the content towards these sessions. So if you don't like it, blame us, don't blame Working Group Chairs. Their session are from Wednesday on. The tutorials are mainly longer session where we have an eye for educational content or ? the plenary is we try to seek those 20?minute, 25 minutes sessions where they are going to be enlightening and also entertaining. We realise it is not a long time to talk but that is also good to have somewhere in the middle to give a glimpse of something new if they have something new to say and in that respect, we also have the lightning talks as part of the plenary, which are shorter talks about more hot topics. BoFs are more informal sessions, sometimes they don't even have this, you know, presentation style over a mike, one person speaking, the rest listening but it's more mingling environments, it depends on the person who is organising the BoF, obviously. Those are scheduled through as evening events.
So, what we try to make sure in the plenary content, obviously as Rob mentioned, this is an operational meeting. We receive and we want to come up with good operational content. Having said that, this community has grown, again as mentioned in the previous talks, it started veryacademic and maybe more diverted on educational or getting ?? learning from each other. Now, we have more operational coordination content coming in, which comes comes sometimes hand in hand with the policy content so we try to balance that out to keep our audience happy in terms of what we want to see and what their interests are. When you talk to each other you find out some of us are coming from academia and some of us are more operational focused, some of us are more on the coordination side so we are a big diverse group and we try to keep each interest in mind when we select content for the plenary part.
Obviously, this contact needs to complement the Working Group agendas, there is more detailed work being done there, and our main, very main focus is to motivate the RIPE meeting attendants and again, really happy with this crowd, this time. I am not going to attach this to the plenary content part, there was lots done by the local workforce, let's say, to bring these crowds in, but it's good to see that the motivation is there for our getting new people into our community, too. And our primary focus, as a Programme Committee, is also towards there.
And we want the education and entertainment, as well of course, needs to be fun. So if you want to join the gang, we still have spaces. You can mail to PC [at] ripe [dot] net. This is important. We need a bio and a statement of interest from you, together with a picture, if possible. We will hold elections on Friday. The Programme Committee is currently composed of eleven individuals, one of us is leaving, I don't know if he is here, Todd, has done a lot for this Programme Committee so a big applause for him, too.
(Applause)
And so, if you are interested, please do send in your interest to us as soon as possible so we put you among the ?? among those who are candidates for the elections. As voters, you need your RIPE access, as Serge mentioned, you have to sign in, to vote for any individual as well as the talks, you need that account activated.
Now if you become the RIPE PC, if you join the group, it comes with cost and benefit, so you may want to think about it before you jump on the boat. There are expectations, obviously. And when we come up with these Programme Committee plenary content, there is work behind it, it's not just, you know, receiving some talks and then just selecting a few. We recruit actually these speakers, we chase after them, often, hassle some of them. But it's all for a good reason. So, consider that in. You will be asked to work, obviously. You will probably need to read, rate and evaluate and comment for 40 to 70 submissions per meeting. They will be presentations of any size, obviously, consider that. And you will be asked to attend three or four conference calls in between meetings. We try to share the pain in terms of the time zone issues, but mainly RIPE NCC who helps us a lot with these conferences, they are obviously on CET, Central European Time we try to keep that as a primer. And in between, the communication is continuously going on within our mailing list with the presenters and with the RIPE NCC, obviously. What you get out of it, basically you get satisfaction. Satisfaction for serving the community that you are part of, and I am not kidding here, seriously I am a good walking example of that, I take a great deal of pleasure for serving this community, which I have been part for a long while and it's a good place to have an active role as such. So this is not only advertising, seriously you may want to think about that, how it's going to make you feel.
The other thing is obviously industry and peer visibility but like I said, it's about serving the community.
So some facts I would like to mention here.
We received 43 plenary talks and 7 BoFs and six tutorials and 12 lightning talks so far as I counted so this adds up like 68, about 70 submissions that we had to deal with this RIPE meeting. Quite a lot compared to the previous events. So, thank you for the interest as well, not only for coming and being here but also submitting content for the meeting itself.
We had a little bit of a hiccup with our submissions system, sorry for that, the people from the RIPE NCC fixed it very quickly but but that was during the call for papers so some of you had to resubmit your submissions, but now we have a solid system.
What we have done, a little more consciously this meeting is that we wanted to public a draft agenda already in March, quite for, two months before the meeting was actually going to be held and the reason for that we were receiving constant feedback that people would like to know more about the content of the meeting so that they could make more concrete travel plans, so the draft agenda was published on March with most of the talks that you see today, and the final agenda was published about a month before the meeting started in ?? on April 13th. We can change this. If it works, is it too early, is it too late, is it good enough time for you guys? Let us know. This is one thing we would like to know from you guys as of the experience. If you want some mic?time this meeting, you can still get some. And that will be ten minute lightning talk. Maybe we still have two free slots for the Friday session, so if you want to submit something you still have time. We recommend five minutes talking, five minutes question and answer for these short talks. They are supposed to be about some current hot topic, recent, that is why we also keep it so late towards the meeting, to evaluate these presentations.
Finally, if you want to work on your presentation skills, like I am saying a bit too many mm?hmms, I learned that, I realise I do that sometimes because I went to tutorial this morning about half how to deliver better presentations. If you have missed that, you can find the slides on the website. It's quite good. I think it's a good balance, having good content as well as you know having the speaker to engage with the audience, so you may want to see if you can improve even better on that area.
And how you can help more. We need help, still. First of all, like I said, you can go for the PC, obviously. There are seats available. You can join the PC and work actively. You can talk to us. Yes, coming back to the slide. These are the people with help ?? Will Hargrave, catch one of us, and if not you would rather write, write to us PC [at] ripe [dot] net. We are looking for suggestions for speakers, and suggestions for content and criticism, maybe, what worked, what didn't worked, let us know and the most important, I say this as the last, Serge is looking at me. Rate the talks, please. As mentioned, we have a functionality on the agenda next to each talk where you can rate the content and how it is delivered. So it's very important for us, this is our lifeline, almost, to see where the interests are lying from the ?? in the RIPE community so we can bring you the RIPE content so it is extremely, extremely important. Thank you.
(Applause)
AUDIENCE SPEAKER: I just wanted to add one thing that, we didn't make clear with rating of the presentations (Serge) the information that is gathered from that will only be for the Programme Committee, it's not that we get the ratings and ah, look what this guy got and that guy got, it's kept for the purposes of the Programme Committee and them only. If you are a presenter and you happen to want to know what sort of feedback you got for your presentation, we are happy to give that information out to you and only you so just be aware that this information won't be made public.
FILIZ YILMAZ: Thank you. One more thing I forgot to mention, you are not only helping the RIPE Programme Committee by rating the talks but you can also win a prize, there are two cheques, 100 worth of Amazon gift cheques. So go for it.
JAN ZORZ: I am a member of Programme Committee, I will take care of housekeeping for this session from now on. I would like to add, we as a Programme Committee, we need information back and we need the feedback from the committee so if you see a guy running around with a sticker that says "PC" on the badge, that does not mean that we are using PCs; that means Programme Committee. Just because people get confused what that PC means. So, welcome to Dublin. Great to see you all here and I would like to invite Olaf Kolkman on stage with a presentation of innovation at the waist. Let's see what Olaf has to say about the innovation. Thank you.
OLAF KOLKMAN: That is quite unexpected being the first, but let's see. Not too many ?? you have a carry?on mike. So, and around RIPE 40 I think I said DNSSEC will be done in six months. So, well, I have been thinking a lot because it's now ten years down the road or even more than a decade down the road and why did I say that? And I think my personal driver is still the same, it's about getting an open, global and trustworthy and innovative Internet, and why didn't that work? That was sort of the question that I was mewing about. And this is ?? I am going to warn you this is going to be a completely non?technical sort of marketing type of presentation. But I had this question: We the Internet community have introduced a number of technologies, such as IPv6 and DNSSEC, have a difficult time being deployed. And the question that comes with that is: Is the network so ossified that we cannot innovate at the waist? So I started to think a little about about innovation and you soon hit if you do a Google query at the work of Edward Rogers, he word a book in 62 called the diffusion of innovation, I think it's now at the fifth edition and is a marketing person. But he was the one that came up with this deployment curve, this S?curve with innovators, early adopters, early majority and late majority and so forth, this type of terminology. And this guy looked at what do people do if they decide to innovate. And this goes for technology, but it also goes for detergents, and washing machines and what have you, so these are relatively, you know, broad set of things.
If you look at the things that people go through when they make a decision to innovate, there are five decision stages: These are them, let's go through them: First, there is the knowledge stage. This is the stage where people are exposed to but don't know much about the innovation at hand. So they have heard something in their social circle and they start to look around. They know of it, they have heard about this IPv6 thing and they have seen the task force or something or they were at a RIPE meeting but they don't know much about it.
The second decision stage is when this individual says, well, I know about this thing but this could actually be worth something to me. So, they will go and seek out persuasion, this is the persuasion stage, where they try to persuade themselves that there is something in it for them. Then comes a decision. Decision stage is where you weigh the benefits and the risks and decide to adapt or reject. You know, this is also the place where you actually have to defend whatever you are going to do, say to a board management and then you have to consider that that upper management might not even be at the knowledge stage, they have to go through this whole cycle themselves again.
If the decision is to reject, then it will take a new persuasion phase in order to reconsider, so new information needs to be brought to that person to make the risk and benefit decision again.
After the decision is made, implementation will happen. And the innovation ?? you know, this is the time where you really have to go through the main pages or look at your vendor's documentation or try and exit and go to the lab. It might be a point where you find out that you lack the information or skills to actually do stuff and you have to invest in this. And the final stage is confirmation. And this is where you basically decide, this thing has value, this detergent actually gives me cleaner laundry. This actually gives me some sort of value. In an engineering environment, this is the place where you move from pilot to production.
If you look at these decision stages, then there are a couple of questions, you know, how does knowledge spread and how are these decisions made? It turns out that knowledge, social networks drive the spread of knowledge. If you sort of speak the same language, if you have shared values, and lack awareness, then knowledge spreads more effectively. Talking to you guys, to fellow engineers, so to speak, it's probably easier to mention IPv6 than talking to CEOs or CTOs in a company that really look at the bottom line. They have a different set of values, and therefore, different awareness. It's fairly important part, that social network that drives the spread of innovation.
Decision types, there are a couple of them, optional, innovation decisions, collective innovation decisions and authority innovation decisions. If you look at those different types, then I don't think that in this environment, if we look, for instance, at innovation issues at the waist RIPE IPv6 or DNSSEC, that the authoritative innovation decision is one that takes place a lot. We collectively don't make that decision. It's more a collective innovation decision or even an optional innovational decision. I am not going to talk a lot about this these types of innovation ?? of this types of decisions.
But, if people make decisions, what are the properties of that innovation that go into the trade?off of their decisions? Well, there is again a bunch of them. Relative advantage, complexity/simplicity, compatibility, observability and trialability. Let's look at them. Complexability, simplicity, if this thing is difficult to use, you probably put it off. Much less likely to adopt it. If you suddenly have to use a different detergent that comes in, that you have to cut pieces off before you can put it in your laundry machine, it's going to be unhand /AOERBGS better to pore powder in a little /PWOFPBLGTS does bring a relative improvement to whatever you do, is your laundry getting cleaner, you have a cheaper operation when you use a different technology. Is the innovation compatable with what you have already created and deployed? Or do I have to change my complete set?up and my whole way of doing things? Those are all kinds of things that make a barrier for a positive decision to deployment higher. Trialability, can you actually try this and walk away from it or once you turn on it, is it ingrained into your infrastructure?
And observability turns out to be relatively high a factor in decision?making as well. Is this thing visible, does it have come cool? Can I show off? There is a lot of people that won't do something if it doesn't show. If it doesn't show that you are better than your neighbour, why would you?
If you talk about relative advantage, value comes in the picture. Met Kaveh's law, the value of the network scales to the log of the number of users time the number of users. The bigger the network, the more value is in that network and you have to overcome that relative value in order to be valuable.
Let's have a look at the Internet. Application, you know this picture, I have made this presentation for a different kind of folk, but Internet hour glass, application at the top, network at the bottom and the IP layers in the middle, in the ?? so IPv4 is the common substrate. This is the place where thank our hour glass where we want to innovate because at the top layer, at the application layer, innovation is relatively easy, that is what made the Internet prosper. People only to make use of that Internet API and IP API could programme against it and of course being a solution in this, most of the people use the http but the success of that Internet thing is that you could innovate on top of the IP layer, you didn't' have to get permission from the lower networks, you could just do your thing. And if you are lucky, the network provided sufficient bandwidth for your application to work.
I claim that at that higher layer knowledge decisions and so on, visibility, relative advantage is high, that made the Internet prosper.
And you know, all kinds of business models work here, proprietor, you can have your freeware, payware, privacy wear, if you think of it, it's basically on yourself, so that works.
All these answers, these decision things are relatively, you know at the positive side if you look at these decisions.
On the lower end of the network, as you know, network and networks, things look a wee bit different but not very much. It's highly competitive, you know know this. The thing is that the innovations decisions that you have to make are in the relatively limited scope and as long as you provide IP to your customers, you can do whatever you want to lower the price of your network transport. You have all the decisions, you can do the ?? you can do ?? you can innovate the fabric within your AS, you can do the compatibility, the observability and the trialability, all of them by yourself, you don't depend on others. And that gives you probably a relative advantage, because if you manage to push your package cheaper than where you are neighbour, you have the ?? the early deployment advantage, because the price of IP is still at a higher end and you just push your packets cheaper than the other guys.
So there is actually, you know, relative advantage and it comes with a price, and a price that is positive on the bottom line.
This, I threw in because it's sort of my ?? sort of my image of why we have a network neutrality problem, those bucks that are made on the upper layer just don't make it through the waist but it's sort of an aside.
Innovation at the waist. Routing security, maybe not that bad, but IPv6 and DNSSEC are two examples of two technologies that we as a community, have a hard time deploying. And that is because, if you look at IPv6, for instance, all these things, relative advantage, complexability, compatibility, trialability, observability, they are not there. There is no relative advantage in deploying IPv6 as the first player. You get introduced to all kinds of complexity, it is not trialable. It's not observable. At least not in the beginning. Besides, you have to cope with the existing value, and the existing value of the IPv4 network is gigantic and that means that at the moment that you introduce IPv6 and in some translation IPv4 to IPv6 translation methodology, and your customer suffer, they want to walk away. It's on you to fix anything that takes the value away from your customers. And that gives you early deployment disadvantage. You are buying innovative gear and you have to debug that stuff and that comes with a lot of unknown risks. That is into the pretty picture. In fact, that is a pretty dark picture. So how do we get this stuff deployed and get DNSSEC deployed? Turn on a magic deployment panel and turn some knobs? This beautiful device is on sale.
How to push these needles. Well, there are some things you can think of. If the relative advantage can be made credible by providing a large enough group to say, we buy into this picture, you might get something done for the group. For the individual you reduce the complexity, increase the relative advantage, enable trialability and make observable the things that they do, they might have an easier decision to make. So, on a group level, what can you do? Well, there is ?? there are mechanisms like regulations or subsidy or market creation that you can use. Sharing the vision of where to go through is something you can do by standardisation and making stuff available in products. There is an open source thing to this. On the other hand, as an individual, the deployment factors by providing good tools, by lowering the cost and providing free software, by subsidy or making observable, by providing add?ons. Trialability and making stuff observable is actually very important. And an example of those are the IPv6 world days, where, by getting a sufficiently large group of people that are interested in deploying IPv6 to line up and say, we are going to do this, and both at the eyeball as well as the content providing stage, people said we are going to do this and at that moment, made observable that IPv6 actually works. So, sharing a message in this case is important. But it was this other, because preaching to the choir is so very easy if it's shared, this has to be apple pie for the people in this room. We need to break out of out of social circus, how do we get this message across to CEOs, to the people who drive businesses? I don't have ?? I do not have the good answers to that, but part of that is to try to adapt your acquaintances to support your values and principles and those principles are needed to, you know, get six billion device on the Net. And keep the open ?? the Internet open, keep it global, allow for innovation and keep it trustworthy or make the impossible, possible, to keep the impossible, possible. That sounds a little bit cryptic but this is sort of the story for IPv6. So these were a bunch of musings I wanted to share. I welcome your questions, your feedback, ideas. And maybe even some discussion, because I believe there are four minutes left.
JAN ZORZ: Thank you, Olaf. I think he deserves an applause.
(Applause)
JAN ZORZ: And I can see the lines forming at the mic. Please state your name and affiliation and ask a question. I think Marco was for a split second faster.
MARCO HOGEWONING: I have been reading up on this as well in the context of IPv6. Not as much question but a suggested follow?up if people are interested in this. There is lass great book by a gentleman called Moore and hysteria ?? hysteria that there is a chasm between the early adopters and innovators and early majority, especially when it comes to high tech innovations and that is quite interesting follow?up read on what Rogers is doing in terms of that gap in the complete detach between us here in the room who are willing to accept new technology and are willing to make it work, and basically there are the majority who just want something that works right away and in a non?disruptive manner. Where do you think in terms of IPv6 and DNSSEC, should we stick with Rogers and accept that sooner or later that gap does not exist, or do you think that we really are looking at that the gap now, where do we stand? Do we fix the early adopter or are we done with the early adopter, and how do we reach out in our group to get the early majority convinced?
OLAF KOLKMAN: I think that what you just said is not ?? that somewhere between early adopters and early majority. In honesty, I don't think we are even beyond the innovater stage here. But that is for the interpretation. I find it very hard to make any assessment of what you just said, but what I think is that what is important if we want to bridge the gap, is that we find the people who can bring the gap. It's probably not us; it's probably the people who are able to communicate the values and communicate the value of a network, an open network in the future. We should help them with that, but talk to your boss. Is what I say here.
MARCO: You are right, that is where are we if we see only one percent has IPv6, are we still really in the innovator phase, and what is the shape that have curve in terms of...
OLAF KOLKMAN: What is important is that in terms of pieces, the building blocks of the technology that we have, because IPv6 is not just one big suite of technology, it's little pieces; there are parts where we are much further down the line, where people have bought in, into the shared vision of big value, but the fact that this thing is an IPv6 stack is an example of a few big decision?makers believing that IPv6 is something that is worth investing. So, if you look at this, for different pieces of the technologies we might be on different pieces of that deployment curve and within institutions that make decisions, people are within their institutions of different pieces of deployment, so that there is a very complex set of dynamics here which I observe but don't have good answers to. This is not my profession.
MARCO: Me neither and I was curious and started looking into this but I will leave it to Benedikt.
Benedikt Stockebrand: IT training consulting. Now, there is a couple ?? I really like the talk mostly because it explains what went wrong when people started to think about IPv6 because they kind of screwed up every possible way they could. There is, however, one thing missing in your presentation and that is basically the question: When do people go to the doctor? When they can't stand the pain any more. So when will people move towards IPv6 if they don't have another choice? And the one thing that hurts management is losing money. So basically, what I found helpful to explain to management that they need to take care about IPv6, is telling them, OK, do you need the Internet, because if you don't, get moving pretty soon now, you will lose your resource Internet because it will have to move anyway.
OLAF KOLKMAN: I think that is right. I think that comes to the argument of does the innovation bring relative advantage. At the moment you can make that point to the top floor, you are done.
Benedikt Stockebrand: Yeah, it's not exactly ?? it's relatively advantage or it's more, you have a significant disadvantage if you move along with the future.
OLAF KOLKMAN: Yes.
RANDY BUSH: Randy Bush, IIJ, eLayer UK, actually. There is something you left out. Many things of course but not ?? a critical one. With DNSSEC, it's forward compatable with DNS. It may be slow to deploy but it's going to deploy. With IPv6, it is incompatible on the wire and there is an alternative: NATS. We have essentially already lost the game. We may pull it back from the brink, but, be prepared to live in a world of NATs.
OLAF KOLKMAN: I am going to bring up this slide again, Randy.
RANDY BUSH: It should be the dark one. And do remember I work for the company that was first deployable of IPv6 commercially in '97.
OLAF KOLKMAN: I am happy you say that because I think it's very possible that we cannot bring it back from that direction, but personally, I think that we have a hard time maintaining these properties when we head into the dark picture, and that is also a reason why I stand here.
JAN ZORZ: Thank you, Olaf, thank you for all the questions. Unfortunately what Randy says it's true, unfortunately, sad but true. So next speech is the title is so long I have to read it, so seeing the past, present and the future, macro trends in networking and the role of software defined networking, probably something about SDN. And it's David Meyer from Brocade.
DAVID MEYER: I changed the title of this but the other title was too long. Before I go going here, this is a great meeting, you guys should give yourself the hand, it's fantastic. Thank you for all for setting this up so well. So this is going to be completely different than what Olaf talked about because I took ?? I have a tend see to take the other approach, I like to look at the science part of it so let's see if we can learn something.
So, my agenda is something like introduction to what I am going to talk about but also I want to talk about architectural features for scalability and evolvability because all those abilities that trialability, compatibility, apparently SDN can do all of that, I am going to question that a little bit and I will show you a little bit of a model if I can get through this for the SDN design space because it's narrowly defined in a lot of people's minds but it's quite of a bigger thing and then a few conclusions and maybe Q and A.
So here is the first thing, this talk is really intended to be controversial and provocative and a little bit sciency, the rule of thumb is you lose 50% of your listers per equation, so I am trying to keep it down.
So here ?? here is the introduction to what I want to talk about. First off there is lots of hyper and open flow SDN, software designed storage and there is hype everywhere around here, everybody knows that. But I was trying to understand what was really going on, is there anything really here and so I went back to these basic architectural principles to take a look at all of this and as many of you know I have been working in the control theory system, biology systems, complexity, quantitative risk engineering and stuff like that for a sometime so I have been, put this together and shook it up and took a look a at what came out. And one thing I can say we need programme attic automation of things like configuration management and a lot of that crosses over into the SDN stuff but a lot of that is already around. And I will just note that part of what I think is the goodness of all of this SDN stuff it's brought ?? being on the Internet space we have always been this the open protocols and all of that but we are moving into open interfaces and API, protocols mostly always have been that way but open source. And this was one of the macro trends that I had talked about but I compressed this down. Along with everything software being macro trend, everything is going to software.
And perhaps it's obvious but scalability and evolvability are key and this goes to what Olaf was just saying in building and operating the Internet and so I am wanted to ask the question: If we are going to do something like SDN, what are even scalability and evolvability and what architectures do we have around that help us with that. What is going on with this. So that is what this talk is about.
So, I don't know how many people know Sean Durane? Anybody. So I was sending my talk and he gives me feedback on it, so I had this bottom line of all of this I really want to ?? want people to see that really what is going to happen is things are going to become a lot more uncertain and volatile, not exactly in our space and I will explain why but in the future, and I will explain why this is the case, how SDN and in the rise of software and all these different roles in general is accelerating this fact and what we might do to take advantage of it and he said forget take advantage of it stuff let's see if we can survive it so maybe that is the right way to think about it.
So what are scalability and evolvability, that is what is at the heart of this? And when I was running had this slide, why do we care, I think it goes without saying we want to be able to scale the network, 6 billion or whatever and we want to be able to evolve and want new features and all that. That said, the way that I am thinking about scalability, it's robustness to change in the size and complexity of a system and that is kind of an intuitive way of thinking about it. Evolvability is robustness to changes in the lineage of things like pieces of software and protocols in our case and our things on longer time scales and you can think of other system futures, reliability is robustness to component failure, efficiency is to resource scarceability and rearranging components, just light in software.
This holds for protocols systems, even Ops, even a lot of things we do on Ops. That is fine but you have got the answer the question: What is robustness? So this particular definition comes from John Doyle and his colleagues attical tech and some other folks but this is pretty good way of thinking about it. If you have a system, the Internet, some property of the Internet, the routing system, whatever, it's sort of bust if it's invariant to a set of perturbations, what would be in the routing system, how you didn't get ?? somehow you survived DS 7007 or something like that. And fragility is kind of the opposite of that, and in reality, fragility is second of a second order effect and I will explain that in a minute but it's basically the acceleration of damage. And interestingly in systems that we care about, the Internet, life on earth, economies, things like that, they have this property that Robust Yet Fragile which is kind of a key thing and what it means you can make a system robust to one set of but become fragile for a different one, completely different in some cases. An example would be if you have high efficiency you don't use a lot of resources, that could cause the system to be unreliable or hard to evolve or and we see this all the time when we are building software or networks. So maybe a picture of this will help. So this ?? the text on this picture is what I just said, but basically, what happens is, when you in these systems when you create robustness, it creates fragilities, in other places in the system, and I will give you an example: When we were ?? I remember when we were first building campus networking, put a subnet in every building and wanted to build robustness to the failure of that interface, we didn't want the hosts inside, building to have to carry routing V VRP or whatever and what did that do? Created robustness to the failure of any one interface but it created a fragility to bugs in the code that managed the Mac address that the router had or where the ?? which router was active, and we had bugs and it happened. And the conjecture here is that this RYF trade?off, if you push robustness some fragility will pop up, that is conjecture to be a hard limit that you can't overcome and there is so much evidence to this, again this is ?? lots of math. One of the thing about this picture is that when people show you graphs, and their graphs responses to things, if it turns out to be concave, what that means really is that the property is fragile, so if you look at where it says fragile here, it's concave. If it's convex, it's robust. And you see that ?? you see that in a lot of things. I mean it's pretty much common.
Some examples of this: So, by the way, this is coming from systems biology, from economics, from social sciences, and from technology. It's kind of all the same in some ways.
So, what is robust? Well efficient and flexible metabolism, what is the fragile part of that. Things can go wrong, you can become obese or get diabetes. What is fragile, micro ECO systems, they can be wiped out easily. Immune system, pretty robust. Information or oughty immune diseases, fragile. Regeneration, renewal at the cellular level, robust if, it goes wrong, fragile, catastrophic. Complex societies, robust, right? But what happens? You get epidemics, wars and things like this. I hope I have the right set of slides here, Jim has a blog on the site about the Syrian Internet outage and it actually is instance of this. And we have events technologies, we know this, nobody has ever seen a bug in a router or anything like that it, wouldn't crash, that never happens.
So basically evolved mechanisms like the Internet or events technology of any kind, they allow for robustness but then they facilitate by virtue of that robustness, novel fragilities elsewhere, it's a property of nature. I need to look at the time.
It took me 20 years to figure all this stuff out so I will try to get it down in 20 minutes.
So, often, often these novel fragilities involve hijacking or exploiting the mechanism that made it robust like cancer or like D?Dos, ampification attacks. These are hard constraints, there is they are theorems and proofs and everybody would go to sleep. What is fragility. Let me geek out on this for just a second. Don't turn off if there is an equation in here. Basically, it's a pretty interesting thing and I started learning about this from these quantity guys, who build these quantitative models who are trying to understand during the collapse of the banking system whether or not ?? they were trying to develop stress tests for banks in the EU Union. And basically what they came up with, what they learned was that this ?? this idea for fragility is really a second order effect and that really has this property that the harm function is non?linear, OK and that is really important and we see that in the network all over the place if we see it that way. And I won't spend a lot of time on this but let me show you what can happen here. If you take something like you have a glass on the table and if you drop it an inch, 100 times, it will be fine, right, but if you take it one time and drop it from 100 times one inch it will be destroyed. So that the response is non?linear, right? And it turns out that nature requires this property because if you think about it, consider jumping off something ?? this thing, you know, I don't know, 30 times, you will be 30 feet, if you jump off 30 feet, something 30 feet one time, you are going to be dead. So it requires this non?linear behaviour. And you know if you think it's too theoretical, we have seen it, ash storms, congestion collapse, are all explained by this.
How many times do people ?? have you said that scale likes ?? we have that all the time because adjacencies turn out to have that scaling property. Another way to look at it in terms of fragility it means that the damage to the network has constant acceleration, second derivative too and again Arp storms, congestion collapse. Anti?fragility is a cool idea from the same guys who were thinking about the fragility stuff and if you look at Jim Cowey's blog on the Syrian Internet, he talks about this it's really quite interesting.
So all of this stuff is finding application in our real lives nowadays. So you look at a systems view of this, the stuff is all heavy tailed so it means that if you want to create robustness in a system like I was telling you, I want to have robust tonnes a failure of one interface, I have got to create complexability, so you need complexability to create robustness but the problem is too much of a good thing and so, it kind of ?? if you are over here, you are good; but if you wind up like over here, it's all pain and you have to get a new job. So, you don't want that. So I am down to ten minutes. All right.
So, we talked about what is robustness and what is it versus complexability but what is complexability? That is another thing. There is so many people, you hear people go that is complexability. What is it? Sheer a definition, you can read this later, I have been giving you guys the sites here but it what it comes down to is structure, the structure that the system has is what its complexity is. So this might be obvious but networks are, they are incredibly express sieve structures, they are graphs, they are extremely common in nature, I mentioned a few of them, but I always told my kids 99% of what you see can be explained in terms of network or a queue and that explains, if you are in a store why wait in line. It comes as no surprise we are studying these to try and understand our network. What are the architectural take aways, why does this even matter? I am kind of racing through this but trying to get to the thing that Olaf was talking about. So we learned that there were fundamental building blocks to these systems and I just talked really briefly about this RYF complexity stuff but, we know that scaleable systems are massively distributed, they are highly layered, we have that. Protocol based architectures we have that, we have redundancy, which is, can be built from degeneracy
Here is the thing that I really found interesting and really clicked in my mind when I was looking at this. So, the biology guys have this idea called the bowtie, and they called it constraints that deconstrain, on the left side of this over here it's just chemicals and stuff like that and it all gets formatted in through this, a knot, and it gets formatted through that and out the outside you get something like ATP which is a standard molecule that all life on earth is built on to, every cell on this so this thing is called a bowtie, I am looking at this and going that is interesting. But you know, I mean it it kind of looks familiar and it's the same idea and these guys were seeing it as horizontal and we see it as vertical but it's same idea. This is a property of a systems that scale, that they have this architecture and this also explains why as Olaf pointed out, why IPv6 is hard to deploy. And Randy pointed out why DNSSEC would be hard to deploy but in here some references again. But if you think about it that is property of scaleable systems.
So here was the SDN thing. This end people were telling us, what happened in computer, it virtualised, standardised, open source software all of that stuff. Well, why don't we just do the same thing in networking, do the same thing. So the question ?? and that is really the thesis, so the question really becomes, oh ?? let me say this thing was populated, character identifies by, it had separation of control and data plan, open interfaces to the data plain and centralised control, at least logically. Is it a good analogy, we were being told that, everybody bought that. So here is the real picture. There is a centralised controller, some packet forwarding hardware and you talk to it and put flows into it, that is more or less what it is. So, what was it on the compute side? Well, it didn't change the architecture, it just opened interfaces. So, I am kind of claiming it doesn't really look like a good analogy even though that is the thesis underlying all this stuff and working for a vendor, they look down when I say this, but the better analogy would be ?? like Linux and running on white box hardware. So I think that is a Bert analogy. I don't think the thesis holds.
Let me say one thing about this. So you have ?? the open flow SDN guys had the centralised controls so what they realised is we knew this all along, you need a distributed system to scale and for resilience, well, it's logically centralised meaning the controller itself was a distributed system. Once you have that, you have all the control plain convergence things and all of the consistency model stuff, so also you have a harder not easier problem. And by the way there is an architectural implication here: If you make fade sharing, the lower band on convergence, this is the physics of it, you have to RTT between the switch and controller and prosets packet on each side of that. So you can't do better than that. We can do better than that right now.
So, another thing is, none of this is new, I mean here is a list of things.
So, five minutes. Let me try this, I don't know if I will have enough time to get this. This was the OpenFlow switch model and the idea here was packet came to the thing, to the switch and the switch model was TCAM, if there is a ROA it applied the ROA and it could either forward it or it would fun to a controller, this was by configuration, up or down, that was it. The problem was it's too simple, you can't do anything with that. That is like an ACL, they were just reusing the ACL TCAM that you have in every router. So this was built. Multi?table, and I will explain a little bit about why but basically there was a explosion plus TCAMs are hot and expensive. If you look at this thing, well, first off, how many paths through the switch, not the network, the switch, it's m?factorial, because N minus choose the second one, you don't want to see that turn up in your analysis of this, it gets worse but I tried to calculate it but it was too hard so I just left it like that.
So it's too complex, it breaks these really interesting reasoning systems, I don't know how many folks follow this stuff but it's being done in colonel and Preston and others, and it had, we had these architectural questions, so the question is, is this really the right abstraction for programmability for what we want to do? I might get through this.
Here is what I think about SDN right now. I see it as this multi?dimensional space, OpenFlow SDN guys thought it was this, complete separation of control and data, plan, OpenFlow and maybe forces and other things like that. But if you move over a little bit, there is something I have been calling control plain SDN so this retains the existing control plain and makes it programmable and so this like the IETF and there are things like PCE, IDLS, vendor SDKs. And if we go way over to the other side, you have OL/SDN and retains the existing control plain and maybe simplified, does it care what the control plain is, it just ?? to the underlay and it may use OpenFlow if they want to programme virtual switches, and that is what they do.
So you have this space, right, that looks like this. So now back to the hour glass. So, let me start at the top. This overlay SDN stuff it's just an application from the network point of view. Control plain SDN is kind of messing around a little bit in the waist but not too much, mostly just with the control plain. The OpenFlow SDN stuff is proposing some kind of new waist but nobody can say where and from my perspective it's too low. So, that is kind of one way to think about all of this. And as I mentioned, in the age of SDN you are going to have to open loop control because even if you have something like ITRS that is open loop control, software and turns of software and power and that is going to cause randomness and uncertainty and volatility, it's all above you.
So what to do? I will just note here, I don't think SDN does anything fundamentally different. Maybe it moves some architectural features around or complexity around but it doesn't do anything fundamentally different so that, Olaf had one of the abilities was ?? had to do with that. But what I would suggest we do as a community is first off, be pretty conservative with the narrow waist, that is what served us so well over the years. Think about these constraints that deconstrain things like IP. And we are pretty good at that. I think in the IETF context we have protected that pretty well. Reuse what we can, that is other thing nature does, trace routes is an example of that, we are pretty good at that.
Expect a lot of uncertainty and volatility coming from above. And we know the networks, we know the networks are RYF complex so we know that there is these ?? these trade?offs being made, even if not being made explicitly, what I would say is if you architecture something for robustness try to figure out what is going to break. Softwares ?? OK. I will skip that. DevOps is something we need to incorporate in our lives and we develop our understanding in the Internet. We change things incrementally, we don't generally try and ?? this is one of the reasons IPv6 is hard to implement, we don't generally try to change the waist all the once. Avoid the talk down. Here is the thing about degeneracy components of the systems are able to do parts of the jobs of other components so that if you want to evolve it, other components can take those ?? that functionality while you are evolving it. So it's ?? you can do redundancy with that but it's more than that. That is a whirlwind tour through this, I think.
I made it down to one minute.
(Applause)
FILIZ YILMAZ: Thank you, David.
DAVID MEYER: Nobody has any questions about any of that? Probably not. No.
FILIZ YILMAZ: Even ?? we decided to take away five minutes from the coffee break but obviously people are hoping for coffee. Just before you leave, please remember rating the talks. Thank you.
