Internet Engineering Task Force 113, 22 Mar 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: IETF113-HOSTSPEAKER-20220322-1545

Description

HOSTSPEAKER meeting session at IETF113
2022/03/22 1545

https://datatracker.ietf.org/meeting/113/proceedings/

A

B

I hope it's. The latest version.

B

No, I just send you a version in the afternoon. You.

C

B

A

And they can upload.

B

It's already here so working stuff.

D

C

D

B

So can I stand to talk or I must.

E

B

E

B

Simply sit here for a day.

B

Or maybe, let's just start because I don't know waste people's time.

B

No, I seem to have echo here.

B

Can you folks hear me.

B

Well, I'm on full screen. I don't know why it's not too. I was going here.

A

So sorry did you simply share the application? Okay, stop sharing and then then go to presentation view, and you should now get back to the technology.

A

A

B

Okay, folks, sorry for the delay.

B

Thank you very much for coming to the host tech talk. This is cheapen xiao from huawei. um There are a million ipv6 talks right. So I first want to answer the question why you should also listen to this one.

B

The reason is that that I think that most ipv6 talks are given by ipv6 supporters and usually the ipv6 supporters. Only tell you the good thing about ipv6 and I think that what's unique in today's presentation is that we are going to talk about a lot of the challenges facing ipv6 and the reason why I feel that this is important and useful.

B

Is that just like you know, nobody is perfect. No technology is perfect. You know every technology have certain problems and in order to get all the people on board to deploy ipv6, we really need to acknowledge the challenges. Certain challenges facing ipv6 document, the challenges um possibly provide a guideline, and I think that uh eric winker kind of like set a good example for us in rfc 1999, because I've seen 1999, basically document all the ipv6 security excels, um provide guidelines.

B

This way, you know people feel more confident that, oh you know, ietf no other excuse, and you know they are solving. This is actually you know, provide more confidence, not less confidence.

B

So with that, um I really want to say that we spent a lot of time presenting uh preparing for these uh presentations and a lot of people provided a lot of uh you know either statistics or provide a local review, and the list is long, and here I really again want to thank eric binker because he provided a lot of you know very detailed feedbacks.

B

So the first part is the momentum.

B

While this is both for the ipv6 supporters and the ipv6 deniers by ipv69, I really have no fixed respect. You know just easier name to refer to the camp.

B

I know that the ipv6, for example, in the nano uh in the nano mailing leaks. There was a big debate this month about ipv6 and some people are complaining that our ipv6 is not easy to deploy, there's no end user demand, etc.

B

While we acknowledge that later here, we want to say that you know for many of these. Are you aware of the ipv6 progress in recent years, because 10 years ago, if you are consuming ipv6, you are more or less justified, because at that time there's almost zero ipv6 users.

B

But now, according to this google stats, we already have 36 percent ipv6 user. um Some people argue that actually, due to the way google count the users, many ipv6 users in china are actually mixed. So if you count e also the users in china, it could be as high as 45 percent based on the etsi study, but while ipv6 go very fast, the the it's very uneven- and this will have a major implication later- that we will talk about so what's driving these ipv6 momentums according to jeff houston.

B

This is you know the gdp per capita. You know which country usually are more enthusiastic in doing ipv6. The user grows like a reliance on ipv for salties and also the competition, but it's not just this. It's not just this, and here we'd like to offer our perspective other. You know strong momentum on ipv6.

B

The first one is that we believe that the loyaltineness of ipv6 value chain is something very important because the ues, the network and the application train is finally ready because 10 years ago, if you know some of these, the content are not already they. Even if operator deployed ipv6, you are not going to get much benefit, but now it's very different, and here we list some of the major milestones, but the slide is busy.

B

So you don't need to read that all I want you to know is that it's summarized in these two blue bars by 2011 ietf largely get the solution already by 2017, the ues, the ues and the big application. Like you know, google netflix are already and this year we believe that there is another important milestone that many people may not realize. Yet is the public cloud support of ipv6 like as aws?

B

These are important. We will talk about why you know this is this may be quite significant and then on the value trend, uh we are seeing that the network is still slightly behind the ue, um the content, and if you are interested, I will not go into detail. If you are interested even later, we can discuss why you know the network is 42 percent, how we computing that etc.

B

um Another point is that I don't know whether ipv6 the denial realized that the price of ipv4 is going up so quickly and specifically the price doubled last year, and if you call it long 2014 it already increased seven times. So, if you the world on ipv4, you may have to spend a lot more money. If you have to buy ip default address or for some operator they are even considering you know, going to ipv6 quickly and while ipv4 you know, price is very high, they can sell their ipv4 address.

B

This is important one and then the public ipv6 weight. This is the kind of the other momentum that we saw.

B

One is this: just now we mentioned that aws ages, support of ipv6. Why? We believe that this is uh maybe very important. We when we look at the operators, you know the big operating six terms, the small and medium enterprise.

B

We find that small and medium enterprise have the least motivation to move to ipv6, because they they have a very you know they don't need mainly address and also they may not have the technical expertise to make the move. So usually you know we. If we look at the sme, you know there's a huge number of smes.

B

They are simply in in ipv4, but we believe that when these you know aws as when they support ipv6 and as the you know, sme go to the cloud you know, uh aws may enable them to move to ipv6 very quickly in the past. They may have to do a lot of work themselves, but now they simply select ipv4 or ipv6 and if they select ipv6, they go ipv6.

B

So we may know today in the v6 working group, there's a discussion.

B

You know how to move the sme and we believe that this may be a very big momentum and then on the iot, you'll probably see the number already there's a lot of iot device and then 5g bring new bl and new opportunity and then obviously there's these innovations on top of ip v6, srv6 and other things, provide better traffic engineering, network programming, etc.

B

So for the people who don't believe in ipv6, I think we really want to. These are some of the recent progress not to mention the. For example, the big government proves you know us government is pushing towards ipv6.

B

Chinese government is focusing, but I guess maybe ietf people don't want government regulations, so I avoid them here and the second part. I think that the first part is more to the ipv69 and the second part is more to the ipv6 supporters.

B

um Here. What we want to say is that for the ipv6 supporters there there really are more challenges than you think, because you know we as the the supporters. Sometimes we feel that, oh you know our work. Ietf work is done. You know these people are not doing ipv6 just because of the lack of the motivation or because they are, they are saying all kinds of skills.

B

You know I am completely ready and what I want to say here is that yeah we are, we are quite ready, but not completely, and let's see so first, this is a survey from every nic. It's done in 2017.

B

You can say that oh 2017 may be too old, but actually you know based on what I hear from other industry forum, if I compare it to some of the ipv6 challenges in in december last year, eating the chinese deck marks and really uh here. I think that it may be difficult for you to to see uh other thing, but I just want to kind of like quickly classify them into four categories.

B

One is motivation. I think that for ipv6, this, like motivation, is always one of the biggest uh xu okay. You know my end user, don't care about ipv6, it's a lot of work, so you know why supply so things like this and ignore this is also a big part. You know, okay, you know, I don't understand ipv6, there are so many new protocols, it's difficult, so I think the big part uh ecosystem is the value trend that we talked about. That ue.

B

Is you know it's my that's my cpu support, ipv6 and the network, the routers, etc, and then technology is what's closely related to ietf that we will focus on today.

B

So, regarding this motivation, knowledge, I think that you know these uh slides look a little deeper. I think that maybe we focus on the the more uh technical part is the technology. Is the technology so that the legacy device is the ipv6 the network management system? I think, generally, this is a field that you know ipv6 network management compared to ipv4.

B

I think that you know there's still some work to do and the oam also may be not as mature. These are the some of the you know, common complaints from the field, but we want to say that, yes, these are issues, but you know the industry largely you know, taking uh action to to deal with them. So, regarding the motivation and knowledge you know, leading operators are sharing that experience and then you know there's many ipv6 forums councils.

B

I think, as far as I know, the uk ipv6 console is is especially you know, active and they're doing a lot of good work and then for the ecosystem. I think it will so clearly. Thank you, know, google amazon, I think microsoft.

B

They really did a lot of work at all. They really did a lot of work to move the industry towards ipv6 and for the last part, for the last part, we are going to talk about the technical challenge. I think that it's the focus of our presentation today so before we talk about the bad things.

B

First, let's uh let's say that you know ietf has done a lot, so you know this is not to to complain. This is not to to complain, but even though that we have done a lot, I think there are remaining challenges remaining challenges, um so the first one. I think that you know to compare the ipv6 performance versus ipv4 uh may be very relevant and more telling, because between ipv6 and ipv4 there are so many things that you can compare. There's so many things that you can compare.

B

For example, neighbor discovery is very different from app, but we believe that you know performance is really the important or maybe the most important. You know uh kpi. So the first one is regarding the round chip time, the latency.

B

So, as you can see, okay, you can see here over the years the the uh ipv6 road trip time is, you know, constantly improving. It's constantly improving, but today it's the old. You know 2.5 milliseconds higher than ipv4, then ipv4, but this is what average this is whatever right. You know later, I will provide you know kind of like the the the going of thought on my own. You know this data and again this is based on jeff houston's status in ap nix.

B

There are you know in the industry. There are some other statistics like linked all have some statics, rather than cc, have statics, etc. That way- or at least I personally believe that maybe ap nick's data is more systematic, so we are so. I think that uh first, is that you know ipv6 and ipv4 is very, very close, but it's slightly higher on a worldwide average scale.

B

So uh I think the very good question to us here is that what are the top three factors leading to this improvement? Why you know over the years why this is improving? And if we understand what are contributing to this improvement, then I think that you know we can provide a guideline for other operators. That, oh you know do this from day one then you know you can you can you know avoid certain things. So this is the logic time.

B

The second one is the packet loss rate and we find that the data here is really inconclusive. We provide you, know two data points here. The first one is from these hindavi. um It reported ipv6 plr actually lower than ipv4, and this is they. They did the management based on one way of management from uh from china, and it says that you know for them. They find out that ipv4 packet loss rate is 0.25 percent um for ipv4 is 0.33, so ipv6 performs slightly better. They also have some. You know.

B

Distribution so in the asia is, is much better. It's all clearly better ipv6 than ipv4 and for other regions they are more or less seeming more or less similar, and then we also get one data point on one angle and then, on the right hand, side we are seeing kind of like a different result.

B

So here we are citing. You know, rfc 7872 by I think, by fernando by zhang lincoln here and now I think recently. uh Eric winker also have the so-called jeff james uh talking about some new management on on these extension headers.

B

But the the long story short the result is that you know in 2017 the package loss rate for for the taxi which fragmentation is seemingly high, 21 percent and even last year it's still eight percent. It improves a lot, but it's still very high um because we heard from many operators we heard from many operators like their ipv for packing loss rate or their packet loss rate is about 0.1 percent and we take it as ipv4, because mainly most traffic is still ipv4, um but you know that's really.

B

I think that we find in the industry, but there is no data on ipv6 packet glossary or at least ap don't provide it. We will find, but we didn't find it. So if anybody have this data, we would love to hear from you. So if we take it from the the first study, ex02.5.

B

Then ipv6 is higher than iphone, but again this is inconclusive. um We we we don't know uh how much of this overall packet loss rate is coming from those fragmentation, extinction, headers, etc. We believe that this would be. uh You know something good to to find out. um We we. We really feel that you know if we can dig into this and find out why you know ipv6 is losing package etc.

B

Again, this will help will help to to improve the performance and make the the life of those deployer a little easier and the third one is the tcp 3d rig. Again we can see that the the blue, the green one is ipv6 and then the other is v4. We can see that over time the ipv6 tcp failure rate is like again constantly improving, but a database on the data uh today, ap nick, it's still, you know, 0.96 percent and ipv4.

B

You know, depending on how you look, it may be like 0.25 percent or some people jeff houston said: oh, it's almost like zero percent, so you know ipv4 is almost successful, but anyhow, I think that the conclusion is the same tcp failure.

B

Ipv6 is higher, so the question is that is it because of the higher packing loss rate of ipv6? I think that this will. Certainly you know if you have a higher ipv6 plr, it's quite likely that the tcp failure rate will also be higher and then because quick is using more and more.

B

In addition to view to measuring tcp failure rates, we also be measuring quick as well. So I think that these are some of the good questions.

B

um Okay, you know me as a supporter, ipv6 supporter, talking about ipv6 challenges here, and some people say that, oh they they may be concerned that okay, you know I come to the conclusion that we shouldn't do ipv6, and I want to tell you that this is not the case because, again that, just now what we're presenting is the worldwide average world wherever it can, you know the average can hide.

B

A lot of things can hide a lot of things, and if we look at the data, you know more carefully, we will see that you know the wall average can be jerked out by a region badly and again. This is the same stats from ap nic you'll, see that you know the whole world.

B

The rtt of ipv6 is slightly higher, but it's largely you know because of you know swing me. If you look at the world, you know american and europe. Actually, you know ipv6 performed better than ipv4, so I I really feel that these data are useful for for communicating with the ipv6 deniers. You know when they say that. Oh it's bad, you can you you. You have the stats to to tell them the work.

B

uh If you look deeper, what's the case and also there are reports that ipv6 can perform better in enterprises, for example, we provide two references here. You know apple at one place is claiming that you know ipv6 can be. You know, 40 better in the in the data center again you know we provided the reference here. You can check out and facebook is also very enthusiastic in in promoting ipv6.

B

So I think the point here, our real point, our real point here is that ipv6 can perform better than ipv4.

B

If, if you, you have a lot of talent, if you can do it, you know correctly, but it's not always easy. This is. This is a complicated machine and if you install it, you know correctly, you operate it correctly. Then you perform better than uh ipv4, but because it's due to its complexity, you may make mistakes and if you make mistake, then it don't.

B

So I think that this is the concern that we want to want to provide.

B

um As a consequence, I think that this is also another thing that I find as a big surprise to me last year, and it may actually be a big surprise to many ipv6 bathrooms is that many operators ipv6 deployment, are actually in overlay and in the underlay in the router et cetera, they're, actually tunneling, ipv6, nprs, etc.

B

So you know, when the operator tell you that, oh I do deal stack, you think that you know their routers are transporting ipv6 traffic natively. You think that their underlay is also dual stack, but that may not be the case. That may not be the case. This is a big surprise for me last year. So, first of all, let's uh let me clarify what's overlay and what underlay I think this is.

B

This is a notion, it's applicable, to operators and, to some extent also applicable to to data center, but it's not applicable to for enterprise if you are not providing a cervix to other people, you really don't have this, like distinction between overlay and underlay, but anyhow for operator overlay is the cervix layer and it involves the the user device and also the mobile gateway of bng or etc. This is your kind of like your service, lady and then underlays your routers etc.

B

Is the the the to send the send the package um because ipv6 is, I think, the the number one reason is to address this address sorties. You know to provide enough address and if to do that, then ipv6 overlay is sufficient, it's sufficient and why operators don't want to you know many operators don't do native. I ipv 6 packaging talk is because if your router is still stacked really dual stack, then you need a a control, plane management, plane data plane for ipv4.

B

You have need a separate plane for ipv6. However, if you just turn on ipv6 packet inside ipv4, then you only need a single. You know data and control management click, so many operators feel that okay, you know. Actually it will be easier for me to maintain just one per play, then two separate play- and there are many of them- are using 6pe sets. So some of the ipv6 veterans thought that operators ipv6 is largely done, but this is really not the case.

B

First, there are many big operators still doing ipv4 and even for those already deployed ipv6 that ipv6 may still just in the overlay, not much in the underlay. So this is again. You know back to the question that we need to make ipv6 perform better. We need to identify the challenges and make ipv6 perform better than ipv4 otherwise. Otherwise you know there are still a big part of operator network can remain in ipv. For all you know there are. There are many consequence so here uh how to make ipv6 perform better than ipv4.

B

There are these answers. The first two came from jeff houston, the first two point that came from jeff houston. He actually have three points, but I had to delete one point, because you know that point received a lot of protest. The deleted point is that he said that you should avoid stateful ipv4 to ipv6 transition and then kind of that you know is closed. For example, six four x, lab etcetera I'll receive a lot of. You know protest from that point, so you know in the end I delete it. um There are other points.

B

The the next five points you know below the below the the white space. The the other six points are. The five points are you know, based on our analysis. First is move content closer to end user. This will improve both ipv4 performance and ipv6 performance, but there are data showing that you know the improvement for ipv6 is actually uh bigger than for ipv4.

B

So if it close the gap or even you know, uh change the change- the wind, there are some explanation and but we don't have the time to go into detail why this is the case. uh Increase link mtu if you can and change the choose, the tcp mss.

B

These are all to avoid trying to avoid recommendation because of fragmentation. If it happened, ego ego caused very high packet loss rate and then there's this like announcing ipv6 route and fee altering jeff houston, provided some example that you know, if you don't do it, you basically provide, for example, that between australia and singapore, ipv4 is australia to singapore directly back to to australia, but for ipv6 turkey goes along australia to singapore, to germany to us and then back to australia, then obviously you are going to have a lot higher.

B

Chip delay and possibly packet loss and then there's this package filtering. I would say that here we provide some some references, some references, but more work probably need to be done, or at least even always. You know, discuss and document these challenges and then provide guidelines provide a guideline on all of these. You know what can make ipv6 better.

B

So this is the second part and the third part is looking forward. As we said, if you look at the wall avalanche, it seems like ipv6 seems to be a little worse, but if you look at, you know deeper. Actually, as we say, if you do ipv6 correctly, ipv6 can perform better than ipv4 just like in america and in europe.

B

So if we look at the the world today, I know that this is a busy slide and you will not be able to tell the company or read the small numbers. I think that there are only two inclusion that I want to get want you to get from this slide. First, is that there are many companies with very high ipv6 uh user percentage. Already, starting from the you know, the top one is 100 ipv6 and even the bottom one is still like 57, it's still 57 percent.

B

um As we said, these are companies that do ipv6 uh correctly and benefit from egg.

B

They can already move to ipv6 only they can already to move to ipv6. Only and brian carpenter told me that you know every time you talk about ipv6 only. You should immediately say that ipv4 as a service over ipv6, because otherwise people will say that oh this guy is a political. How how? How can you be? You know? Ipv4 will stay for many many years. They thought that ipv6 only means like no ipv4, but it's not the case.

B

It just means that you know ipv4 will be long as a cervix over ipv6, but it will disappear from the infrastructure layer. So it's the the time to to do so, and here you know, will be a little bit more specific. So at least from an operator perspective you know facebook already done it. I think the hyper scale, the hyper scale, have a lot of talent. They probably have you know the best talent in the industry, and they have no problem.

B

You know they can they they they they they can do uh whatever, but for some operators it may not be so clear, so we're kind of, like you know, draw some picture.

B

um I think the mobile one we kind of like use four six four x, a lot as an example and the first one is a little bit like uh you know quite dislike, or you know one of those solutions, but the the real.

F

B

Is that I think that ipv6 only at the moment in the industry for some people, it doesn't mean that ipv4 as a cervix? Again, you know it's just in the overlay, and here we want to make sure that you know by ipv6. Only we not only mean overlay, we really mean you know underlay and the ipv4 the package will be tunneling inside ipv6 will be finally inside ipv6. We believe that this will further improve the performance of ipv6 and make it better than ipv4.

B

So this is largely the the kind of like the last slide before summary is that we, the difference between these one and the ipv6 overlaid, is that we believe that you know the. If you do do the today's way, ipv6 overlay plus ipv4 underlay, you are basically you. You have no exit strategy because you, your your underlay, is stuck even though that your ipv6 traffic is increasing.

B

But you know you, you don't change your underlay and if you always have you know two protocol, but even though, if you tunnel, your ipv6 traffic ipv4 over ipv6, because all ipv4 is, is becoming a smaller and smaller percentage, then you know one day you can finally just purely ip purely ipv6 and really these require solving some of the challenges that we talked about in this presentation.

B

um Really we need the the health of ietf and the leading operators. We really need. First, we need to document all the challenges and provide guidelines. So this brings me to the summary again: we want to emphasize with our feelings that it's like a sophisticated machine.

B

Only if you can install and operate it well, it will be better than ipv4. Otherwise it can be worse. Then it can be worse than ipv4, and in this aspect we really need. You know, ietf, to kind of like provide kind of like better user guideline, to make sure that you know uh people will have a higher chance of you know, install and operate it correctly um these. If we do this, then we can.

B

We can, you know, accelerate the move to ipv6 and ipv6 only and here we will specifically list some topics that we hope to get your contribution.

B

I I will pause here and let you uh a few seconds to read it. I also provide my contact information, so we love to to hear from you.

B

Thank you very much.

B

I think you don't usually get a approach in ietf, so.

F

B

Feel really honored.

B

Any any questions comments.

G

I cut in front of you last time you go ahead.

B

Do you want to speak.

F

I hope you guys can hear me so I have.

C

F

um You talked about uh there having been some work done to measure the performance improvements um and that performance improvement for ipv6 being larger than the performance improvement for ipv4. When you talk about uh dealing with keep edge or deep in the network deployed content key. So I was just wondering if you have any uh study or reference material that I can look at.

F

uh You know, because that is an area of great interest for me and the organization that I work for and uh we'd love to we're already kind of semi dual stack, but we deploy in a lot of networks which are till to this day large. uh You know consumer facing isps that are ipv4 only so we'd love to be able to show some statistics, and you know, try to help move the needle with these. uh You know medium mid-sized isps, who you know. I have just not got onto the ipv6 bandwagon.

B

uh Okay, I I think we should probably you know, take this discussion offline because uh it will it one is, it will take a lot of time and a second you know, I cannot remember you know all the all the you know the the the reference, so I can send you the reference afterwards.

F

I I will definitely email you and we'll we'll take this up offline, but then um you know it it. This is a very interesting data point to have, because we, as service providers who are going into these deep network locations, we would love to have some.

F

uh You know ammunition so that when we go to these providers, who are reluctant to you know uh embrace ipv6, we have something that we can show them and say it's in your interest to do this, even though uh you know you don't mean you may not see a pressing commercial need to do it right now, but hey here's a case for a performance improvement for your customers that should uh you know, help you make the decision to go ahead and do this.

F

Okay, thank you. Rajiv.

H

uh Journaling over google, as an enterprise operator, who spent some time operating dual stack and v6 only network. I have an impression that my experience every single case when you see v6 performance worse than before, is because of api balls, because v6 network is broken. Nobody monitoring could because of happy. Eyeballs users are not complaining and it stays like that until you turn off ipv4.

H

So I'm quite sure it's not something in the protocol, not something in the connectivity. It's just bad operational practices.

B

Okay, yes, I think that that's that's kind of what we are saying that you know the machine itself is fine, but you know: if you don't operate it correctly, then it has problems.

I

Hi this is barry lieber. Can you go back to the slide where you show the performance, differences in different regions and I'm curious on that yeah? Just you just passed. It now go back the other way.

I

Next, one that one um do you have any sense of why oceania is so bad compared with the other regions.

B

I I don't okay, because I think that you know this is a slide that I newly added and I today I went to the the ap unique data and copy this. I think that this, but I think that it's a good question that we can probably discuss together with jeff houston.

I

J

Hi mike ackerman, first off, I want to say that you covered a lot of information. This is really good stuff. I hope you'll come back and do it again, it's almost inspirational. Thank you. um The uh first question I want to ask: maybe somebody already did: can we get copies of your slides.

C

J

Fantastic uh and then I just wanted to ask one specific question: I got your email, I got a whole bunch more. So beware, um you had a slide that said, and it's right up there. It's uh that ipv6 performs better in enterprises. I'm one of those. Can you? Is there a short answer as to why or how.

B

We see it in this, uh for example. Let me let me take facebook as a example. You know facebook, we we provide here. You know this link to a youtube video and in this youtube, video. I believe that the person's name is taught uh well, it's a a facebook person talking about uh ipv6 his ipv6 performance is better than ipv4. I I remember using a nano meeting or in some kind of meeting, and then there are people flown from the audience asked the same question that you just asked yeah.

B

We want to know that those people, I think that you know uh they provide some kind of I. I will not say that it's it's a very direct answer. They say well, for example, if you do ipv6, then you have no net, but really nat will introduce only a very small difference, so ipv6 is better. Is it really because of ned or because of some other reasons we didn't get a a very clear answer from that uh youtube detail. I think we have the reference.

H

J

B

Okay, you can take a look hoping.

J

There was some great generic answer, but it could be application or situational specific and that that's certainly understandable, but.

B

Yes, we actually, we really, I think, like these are the questions that we should dig into, and this is why we are doing this and actually maybe maybe jane also, you know, have some comments. Do you, for example, in your google network? Do you feel that ipv6 perform better or you know if it performs better? Why.

J

I definitely wrote down what she just said that when you get to ipv6 only- and that makes a lot of sense- you can see where the economies the scale will start to add up and give you better performance, maybe even holistically. So I'm just looking for any answers anywhere. I can get it, and this was great. Thank you again.

B

B

So we are already 5 30 and if there are no more questions, I really want to.

B

Rajiv, you have another question.

F

Quick question: um if it's not too late,.

B

Go ahead, rajiv.

F

B

F

Yeah, so this is again relating to the difference. um So, while in this table, you're kind of showing the average uh difference in the rtts uh and and yes based on these numbers oceania, looks like it's a massive outlier with a much larger average.

F

But uh I probably am not as surprised by that, because it's quite likely that the overall total rtt average for connections in oceania are probably going to be a whole lot higher because of the fact that it has very limited uh capabilities of serving traffic locally and most traffic originating from there is reaching out to some other region of the world um in order to get served, which is not the case with maybe uh america's uh or europe or for that matter, even asia, because a lot of that traffic will be within the region.

F

Okay, so a lot of asia is going to be served out of china or india or singapore, and you know, doesn't really have to go outside the region. So when the total uh connection rtts are in the range of 40 or 50 milliseconds, their 10 difference turns out to be four milliseconds perfect with oceania, if they're, if the average of their connections are mostly going to be in the 300 and 400 millisecond rtt range, because they're all going outside of the oceania region to get served.

F

A 10 differential in performance would still look like a very big difference over here. So I'm my question is: have you looked at that as a data point? While doing this analysis.

B

uh The the answer is no, as I said you know this slide, I added today, you know upon the request of a colleague, but I I can offer a observation is that you know if you know jeff houston, how he how he collects. His statistics he's basically partnering with google, and you know uh uh putting a kind of like an ad on the you know, chrome browser, and then this chrome browser will contact his. You know, he's he's he's a prober, let's say, and then um because this is sampling, this is stinkling.

B

So I do the the observation that you know at different time. These can you know, for example, tomorrow, if you go to you, know jeff houston's side and then take another snapshot. The number can be quite different, so I think that this is. This is point number one this you cannot definitely not. uh You know, draw a conclusion that you know osun is always is always bad. So again you know all of these. You know uh every statistics.

B

You must also look at each flow. uh You know kind of like take it with a grain of salt with a grain of salt. You know really tomorrow we when we look at the the one and just now, when you talk about the the maybe you know, oceania is far away from kang tang. But again we are here. We are talking about the difference between ipv4 and ipv6, and the long distance you know is a burden for both ipv4 and ipv6.

B

So, arguably you know it shouldn't create a lot of difference, but other than that. Really I'm sorry, you know I I. I really feel that that for ipv6 performance there are a lot of analysis is worth doing, because we really you know. No, I would say that we know very little and if we we just support ipv6, and we don't face these challenges- they will not just go away.

B

So I think this is why you know I I I like to do this talk and kind of like expose these challenges, and so that you know we can, you know, do something together.

F

Thank you that was very informative.

B

Well, uh let me thank all of you very much for spending the afternoon with us. I really appreciate it and you got my contact information and we really love to to hear from you so again. Bye now.

D

Thank you, so you.

D