Michel Py wrote: Do you measure what is happening on private interconnects ? MMR traffic ?
Job Snijders wrote : Yes, looking at stats at NTT (a network which basically is only private interconnects), I see a similar pattern as we observe at AMS-IX. I'll see what detalis I can share.
I have to admit that your figures are a bit of a surprise; 2.5% IPv6 average. Thanks for sharing.
It would be nice if more players would share a normalised overview of IPv4 vs IPv6 percentages, just like AMS-IX does.
Indeed.
I find it hard to believe that two networks would end up exchanging IPv6 traffic over private connections, and at the same time keep IPv4 traffic on public IXPs or transit. That doesn't seem to align with the usual economic or security drivers behind peering. Of course we can't exclude the possiblity this happens, but I am not aware of anyone who explicitly configured things to be that way.
I was not aware of any either, I thought in Europe it could have been different. In the US, where would the traffic between Verizon wireless (heavy IPv6) and Google (IPv6 enabled) go ? In multiple MMRs / private interconnects ?
I'm beginning to suspect that the "there is lots of IPv6 traffic!" some folks report on is mostly between handsets (strictly controlled by the mobile provider) and a select few Big Content on-net cache devices.
Indeed. Just take Verizon wireless out and half of the IPv6 traffic disappears. Well, not half but certainly a sizable chunk. Google confirms this : https://www.google.com/intl/en/ipv6/statistics.html If you zoom in, you can see a clear weekly pattern of about 5%, which is clearly that at the office people use their office computer to Google during the week, and during the weekend thy use their mobile or their home ISP, Comcast being IPv6 heavy contributing to that. Same thing happens at new year : everyone is at home, so the IPv6 percentage is higher. Can you zoom your AMS-IX graphs so we can see if you have the same phenomenon ? A monthly IPv6 graph with the top of the graph being 3% ? Just right there, we can see clearly that if Google were to go IPv6-only, they would lose 1/6th of their traffic. (that is, IF the top of the graph was 100% IPv6, which it is not. If Google were to go IPv6-only today, they would lose 70% of their traffic). I suspect that the people in charge have made that analysis and that they are not going to lose that much of a customer base, especially when the base in question is business.
Even if we consider such an intranet IPv6 deployment part of the big-I Internet, it doesn't strike me as healthy.
It's why I call it a niche market. Only on environments that are completely controlled by the provider, where the user has no choice (and does not have a clue, anyway). It's an island. The bar is on the beach, and they keep it well stocked (CDN cache) so the booze flows in abundance, but it does not get out much. Healthy ? depends who you are. For them, I think it is. They have the customer completely locked.
I posit: the further an IP packet has to travel, the less likely it is to be an IPv6 packet.
+1 Michel.