Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!


BLACK FRIDAY REVIVAL ★ FIRST 15 BUYERS ★ 30% FOREVER DISCOUNT ★ 16 CORE DEDIS ★ DOUBLE RAM DOUBLE BW - Page 2
New on LowEndTalk? Please Register and read our Community Rules.

All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.

BLACK FRIDAY REVIVAL ★ FIRST 15 BUYERS ★ 30% FOREVER DISCOUNT ★ 16 CORE DEDIS ★ DOUBLE RAM DOUBLE BW

2»

Comments

  • Daniel15Daniel15 Veteran
    edited March 2021

    @BrianHarrison said: Their IPv6 traffic flows are all represented in the SIX traffic stats that I shared earlier.

    I could be wrong but I imagine a lot of cross-server / cross-DC traffic flows through SIX and similar Internet exchanges, which would have different patterns to traffic from regular end users. Perhaps SIX's IPv6 connectivity is not very good, or some people are only peering with SIX for IPv4?

    Akamai are huge and their stats (https://www.akamai.com/us/en/resources/our-thinking/state-of-the-internet-report/state-of-the-internet-ipv6-adoption-visualization.jsp) are roughly the same as Google's, around 44% IPv6 usage in the US. Cisco's data (https://6lab.cisco.com/stats/) shows similar numbers too. So, I think the SIX data is far lower than other sources.

  • BrianHarrisonBrianHarrison Member, Patron Provider
    edited March 2021

    @Daniel15 said:
    Akamai are huge and their stats (https://www.akamai.com/us/en/resources/our-thinking/state-of-the-internet-report/state-of-the-internet-ipv6-adoption-visualization.jsp) are roughly the same as Google's, around 44% IPv6 usage in the US. Cisco's data (https://6lab.cisco.com/stats/) shows similar numbers too. So, I think the SIX data is far lower than other sources.

    The pages you've linked actually show IPv6 adoption -- not IPv6 usage. There's a big difference between the two.

    I could be wrong but I imagine a lot of cross-server / cross-DC traffic flows through SIX and similar Internet exchanges, which would have different patterns to traffic from regular end users.

    That's a good thought, but unfortunately that's not the case either. There are plenty of consumer-serving telecoms on the SIX: T-Mobile, Charter, Telus, Bell, LocalTel, BendTel, China Mobile, Frontier, Shaw Broadband, Sky Fiber, SK Broadband, Telekomunikasi Indonesia and plenty others I missed. Lots of content providers on the SIX too (such as Facebook, Twitch, etc) whose traffic (the vast majority of it) goes directly to end-users.

    It's even more so the case in Europe with the AMS-IX and other major IXPs. In Europe, traffic exchange and peering is more centralized than it is in the USA and the big IXPs handle huge traffic volumes. Those IXPs, like the AMS-IX have similar low IPv6 percentages. To be clear though, you don't necessarily have to be a participant on an IXP for your traffic to end up flowing over an IXP.

    Lastly, if you look at the traffic graphs themselves, you see the clear rhythm of human activity in the peaks and troughs. The lowest traffic levels occur at 4AM when everyone is asleep, early morning peak just after 9AM when everyone is getting to work and the overall peak in the evening at 8PM-9PM when everyone is watching some TV or browsing the web before getting ready for bed. Clear signs of user-driven activity. The peaks and troughs are more significant on the AMS-IX than the SIX (indicating more user-driven traffic on the AMS-IX), but they exist on both.

    No matter how you cut it, the data shows that IPv6 usage is significantly weaker than one might think.

  • @BrianHarrison said: Lots of content providers on the SIX too (such as Facebook, Twitch, etc) whose traffic (the vast majority of it) goes directly to end-users.

    Do you know whether this chart includes completely private interconnects between major players, instead of like a 40G to the IX?

    TMobile traffic is ~95% IPv6 to Facebook, Comcast is 80-90% IPv6 to Facebook. IPv4 is minority.

    Thanked by 2yoursunny Daniel15
  • Some telcos do support ipv6 and embrace it even. By default you would be using ipv6 which is the reason why you see these US telcos with a lot of ipv6 traffic.

    Then you have telcos/ISPs without ipv6 support at all. And some who do support ipv6 but by default ipv6 does not work and you either have to change AP settings or call them up to enable ipv6. You still have people around the world who don't have ipv6 and ipv6 has not proven to be faster than ipv4 in real-world usage. Some services outright do not support ipv6 or it's just not a good idea to use ipv6. For example, some games don't support ipv6. Would you risk sending emails on ipv6? ipv6 still has a long way to go

  • so google have data that show ipv6 adoption and usage are high ?

    did someone mention google also member of SIX ?

    but google cloud is listed in hall of shame ?

  • BrianHarrisonBrianHarrison Member, Patron Provider
    edited March 2021

    @aiden1 said:

    @BrianHarrison said: Lots of content providers on the SIX too (such as Facebook, Twitch, etc) whose traffic (the vast majority of it) goes directly to end-users.

    Do you know whether this chart includes completely private interconnects between major players, instead of like a 40G to the IX?

    Well you said it yourself, those interconnects are private and, generally speaking, there is no public information available on them.

    You may be missing my point though. I'm not arguing that IXPs comprise 100% of all traffic flows to and from Facebook, T-Mobile, Google, etc. There is certainly tons of IPv6 traffic (and IPv4) that occurs off of IXPs.

    My point is that, due to the sheer size of traffic flows on the IXPs (peak traffic on AMS-IX is 10Tbps) and the significant representation of the aforementioned ASNs on those IXPs, we can therefore reasonably assume that IXP traffic statistics offer a reasonably accurate representative sample of relative IPv4 and IPv6 traffic flows more broadly across the Internet. In other words, there is no better publicly available data with which to form conclusions about the relative share of IPv4 traffic vs IPv6 traffic across the broader Internet.

    It seems some folks are having a really hard time accepting that IPv6 usage is a tiny percentage of total traffic flows and that the percentage, at least by some measures, has actually been in decline over the past few years (decreasing in relative terms, increasing in absolute terms).

  • DataGizmosDataGizmos Member
    edited March 2021

    Remove the top 10,20,100 sites and their IPv6 traffic flows- and what do you have left? Probably not a lot. These top sites certainly have different arrangements with ISPs, Interconnects etc to move traffic in efficient patterns so it never even hits the full internet. Example - netflix puts servers within ISP networks etc. So t mobile can travel IPv6 and hit a box within its network and never need IPv4 or use IPv6 through an interconnect thats public. So your numbers are going to be skewed- showing sky high IPv6 internally to t mobile and the large sites with unique arrangements. But- that has no bearing on the 100 million tiny sites in the world. When you remove the large data heavy, often visited sited (FAANG companies etc) they make up a large percentage of netowrk traffic- so not much left and whats left, often doesnt use ipv6

    Thanked by 1BrianHarrison
  • @BrianHarrison said:

    That's a good thought, but unfortunately that's not the case either. There are plenty of consumer-serving telecoms on the SIX: T-Mobile, Charter, Telus, Bell, LocalTel, BendTel, China Mobile, Frontier, Shaw Broadband,

    I imagine Telus and Shaw are bringing down the average. They've got no incentive when they have IPV4 up the wazoo. I'm curious what Windows firewall does when they click home connection when prompted with a routeable IPv6.

    Lastly, if you look at the traffic graphs themselves, you see the clear rhythm of human activity in the peaks and troughs. The lowest traffic levels occur at 4AM when everyone is asleep, early morning peak just after 9AM when everyone is getting to work and the overall peak in the evening at 8PM-9PM when everyone is watching some TV or browsing the web before getting ready for bed. Clear signs of user-driven activity.

    Uh, who is "everyone" given the different timezones? There's 5+ 8-9PM's in North America with an unequal balance of people. Do you mean like a 1000 mile radius?

  • BrianHarrisonBrianHarrison Member, Patron Provider
    edited March 2021

    @TimboJones said:
    I imagine Telus and Shaw are bringing down the average.

    You're probably right, but there are also others who are bringing up the average, like T-Mobile and Facebook :) It all averages out when you have a very large sample size (i.e., traffic stats from major IXPs).

    Uh, who is "everyone" given the different timezones? There's 5+ 8-9PM's in North America with an unequal balance of people. Do you mean like a 1000 mile radius?

    If a user is located closer to an IXP, then, generally speaking, the likelihood is higher that their traffic will flow through that IXP. So for the SIX, the "average" user would be on PST/PDT. For the AMS-IX, the "average" user would be on CET. In both cases, you'll find that the traffic graphs match the approximate 4AM, 9AM, 8PM-9PM pattern that I outlined earlier.

    Thanked by 1TimboJones
  • BrianHarrisonBrianHarrison Member, Patron Provider

    @DataGizmos said:
    But- that has no bearing on the 100 million tiny sites in the world.

    Yep, you're absolutely right. And more to that point, those 100 million tiny websites are more likely to be our customers. Our target market isn't exactly cell phone users accessing Facebook or Netflix (over IPv6).

  • Order #2662143793, hopefully the service is as good as the price :)

  • BrianHarrisonBrianHarrison Member, Patron Provider

    @ankursharma8715 said:
    Order #2662143793, hopefully the service is as good as the price :)

    We hope so too! ;) Glad to have you aboard!

  • vernorvernor Member
    edited March 2021

    I bought the "RepriseEDED Special". The order is in the process for the time being, and once I get the server, I will edit this post to share my experience. However, I must mention that Brian is very assistive, and thank you for offering such a good price!

    --

    I got the server within 2 hours. That's a great performance. Your customer service specialist works awesome! He handled all of my problems and assisted me in every step. I will not comment on the price as it is magnificent. Thank you again.

    Thanked by 1BrianHarrison
  • BrianHarrisonBrianHarrison Member, Patron Provider
    edited March 2021

    @vernor said:
    I bought the "RepriseEDED Special". The order is in the process for the time being, and once I get the server, I will edit this post to share my experience. However, I must mention that Brian is very assistive, and thank you for offering such a good price!

    --

    I got the server within 2 hours. That's a great performance. Your customer service specialist works awesome! He handled all of my problems and assisted me in every step. I will not comment on the price as it is magnificent. Thank you again.

    I'm glad to hear things are working out well! Thank you for your business!

  • WilliamWilliam Member
    edited April 2021

    @BrianHarrison said: That's a good thought, but unfortunately that's not the case either. There are plenty of consumer-serving telecoms on the SIX: T-Mobile

    DTAG does not peer. They are only on exchanges for PNIs. AT&T also does not peer. Comcast enforces ratios. Verizon is a Tier1 and does not peer at all.

    Your SIX data is thus absolutely useless - Eyeball networks use DE-CIX/AMS-IX and PNIs, not exchanges.

    A path from Akamai to Tmobile/DTAG or AT&T will never cross an exchange, and it will be v6 end to end.

    @BrianHarrison said: If a user is located closer to an IXP, then, generally speaking, the likelihood is higher that their traffic will flow through that IXP. So for the SIX, the "average" user would be on PST/PDT.

    Entirely wrong. I have an exchange like 10km away from my house, yet my ISP will route me to Zagreb and back. I lived in Vienna, and UPC will route through Hungary happily to reach a DC in Vienna.

    Ultra large carriers like Tmobile US are not using local exchanges at all, its way too much small sites to maintain - They backhaul to Network POPs in CA/NV/TX/NY (secondary WA, NM) primarily.

    This is also needed as the routing complexity of adding an exchange for 5% of users in CA while routing your entire other network around it is just pointless and requires 50% more hardware (and let me tell you a MX960 is nowhere near cheap.).

    @BrianHarrison said: For the AMS-IX, the "average" user would be on CET

    Bro, AMS-IX is way more than an EU exchange. Same as DE-CIX a LOT (at times 50%+) is ASIAN and AFRICAN. The timezone is fairly irrelevant.

    By now many, if not a majority of, DE-CIX members are not even incorporated or based in EU.

  • BrianHarrisonBrianHarrison Member, Patron Provider
    edited April 2021

    @William

    I respect your opinions and knowledge, but I generally disagree with your response.

    DTAG does not peer. They are only on exchanges for PNIs.

    T-Mobile does peer on the SIX. They even advertise prefixes on the public route servers. I'm not familiar with their peering policies on other exchanges.

    Your SIX data is thus absolutely useless.

    I disagree. I have explained in my previous posts that plenty of eyeball traffic flows over the SIX. But, for the sake of argument, even if I accept your point and we agree that the SIX data is useless, then we can go ahead and use AMS-IX data. AMS-IX statistics show the same low ratio of IPv6 traffic that the SIX data does.

    A path from Akamai to Tmobile/DTAG or AT&T will never cross an exchange, and it will be v6 end to end.

    As I've explained in my previous posts, we could sit here all day and point out one anecdotal example after another of IPv4 and IPv6 traffic that bypasses exchanges. However none of that would change the fact that the large and diverse traffic volumes on the exchanges allow them to serve as fairly accurate representative samples of IPv4 and IPv6 traffic ratios.

    Entirely wrong. I have an exchange like 10km away from my house, yet my ISP will route me to Zagreb and back. I lived in Vienna, and UPC will route through Hungary happily to reach a DC in Vienna.

    I did not say "if you live near a a major IXP, then your traffic is guaranteed to flow through it." I did say that your traffic is more likely to flow through it than if you were located very far away -- which is true.

    Bro, AMS-IX is way more than an EU exchange. Same as DE-CIX a LOT (at times 50%+) is ASIAN and AFRICAN. The timezone is fairly irrelevant.

    I'm talking about averages. Of course AMS-IX handles traffic from other timezones, regions, etc but the average timezone of end-users whose traffic flows through the exchange does roughly equal CET. This is clearly shown in the AMS-IX traffic graphs where the peaks and troughs align with a typical end-user activity in the CET timezone.

    By now many, if not a majority of, DE-CIX members are not even incorporated or based in EU.

    I'm not talking about the timezone of corporate headquarters, I'm talking about the average timezone of end-users whose traffic flows through the exchange.

Sign In or Register to comment.