RE: IPv6 traffic stats (was: Re: Last Call: draft-irtf-asrg-dnsbl (DNS Blacklists and Whitelists))

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



David mentioned:

#For the record: 
#
#It seems that arbornetworks estimates are extremely low to the point
#where one has to ask whether there were other issues that caused such
#a low estimate.
#
#There is no question that IPv6 traffic is quite low in the Internet.
#However, many other reports that I have seen recently measure multiple
#orders of magnitude more IPv6 traffic (for an easily accesible example
#see: http://www.ams-ix.net/technical/stats/sflow/)

The Ether Type graph on the AMS-IX page indicates that IPv6 is on average 
1/10th of 1% all the traffic they measure, and looking at the associated 
RRDtool graphs, that works out to be ~800Mbits/second.

A sustained ~800 Mbits/second is certainly nothing to sneeze at, and
everyone who has worked hard to encourage IPv6 deployment deserves many
kudos. Progress is happening!

On the other hand, just to put this in context and to pick on an 
application I'm somewhat familiar with, a single full-ish Usenet news 
feed is now in excess of 3TByte/day (see the daily volume stats quoted 
at http://en.wikipedia.org/wiki/Usenet ). Just two or three full-ish
Usenet News feeds over IPv6, if done across AMS-IX, would account 
for most of that 800Mbps traffic load (assuming that Usenet is what
was making up most of that traffic, an assertion that I'm explictly
NOT making). My point? It is possible that the IP transport choices
of just a few cooperating server administrators might (at least
hypothetically) account for virtually all the observed growth in 
AMS-IX IPv6 traffic. 

As to why the AMS-IX number might differ from Arbor's statistic, we 
know that traffic at exchange points may have a dramatically different 
composition than traffic measured elsewhere, due in part to the 
economics of that environment. 

E.g., continuing to pick on poor old Usenet, people may be willing to 
exchange Usenet feeds across a settlement-free peering point while 
they might NOT be willing to exchange Usenet feeds that required 
(comparatively expensive) transit bandwidth. Those sort of economic 
choices mean that it is risky to extrapolate Internet-wide traffic 
statistics from the somewhat atypical settlement-free peering 
environment.

But what sort of growth pattern do we actually see at 
http://www.ams-ix.net/technical/stats/sflow/ ?

That graph *isn't* growing in the characteristic "stair step" pattern 
one might expect if you were to suddenly flopping full news feeds over 
onto IPv6. The growth we see there is much more consistent with what you 
might find from growth in end user traffic (which could be dominated by 
web, or P2P, or flash videos or scientists ftp'ing large data sets, or
yes, even email, who knows, since there's no way to definitively know 
w/o doing deep packet inspection, which I doubt would be possible in 
this case). 

So to bring this post to a close, I continue to believe that IPv6
traffic, at least IPv6 email traffic, remains very, very low, to 
the point where, as I've previous mentioned, it just hasn't justified
DNS block list operator attention in any material way (love to hear
about any counter examples, BTW).

Regards,

Joe St Sauver (joe@xxxxxxxxxxxxxxxxxx)
http://www.uoregon.edu/~joe/
Disclaimer: all opinions strictly my own. 
_______________________________________________

Ietf@xxxxxxxx
https://www.ietf.org/mailman/listinfo/ietf

[Index of Archives]     [IETF Annoucements]     [IETF]     [IP Storage]     [Yosemite News]     [Linux SCTP]     [Linux Newbies]     [Fedora Users]