We've had one presentation on the "unfairness" of p2p traffic, which (the presenter says) will eventually swamp us. Then just now, we had the presentation & subsequent discussion re: ipv6 adoption. Just wondering: what if we gave ipv6 traffic "mucho priority" over ipv4 traffic, then tell our user communities that ipv6 provides a better quality network experience, including (hopefully) faster page loads, & lower video game pings? With such policies in place, folks wouldn't want to stay with the "old, slow" v4 traffic...and could be a significant selling point. After all, if most p2p traffic is v4, prioritizing ipv6 (as a general concept) should improve the user experience. Anyway, was just an idea, please pardon me if this has been discussed before, or sounds nutty... Thanks, -Scott
Scott Doty wrote:
After all, if most p2p traffic is v4, prioritizing ipv6 (as a general concept) should improve the user experience.
How long do you think it will take for the P2P software authors to transition over to IPv6? I'll bet that P2P users will be a lot more likely to use IPv6 over Aunt May checking her email once a day. Niall. -- The virus contained in this message was not detected.
Actually, I seem to recall some postings to the list stating that many of the popular bittorrent clients already do IPv6 if available. So that would seem to be a good recipe for allowing P2P users to prioritize ahead of regular traffic. - S -----Original Message----- From: Niall Donegan [mailto:niall@moybella.net] Sent: Tuesday, October 14, 2008 3:56 PM To: Scott Doty Cc: nanog@merit.edu Subject: Re: spurring transition to ipv6 -- make it faster Scott Doty wrote:
After all, if most p2p traffic is v4, prioritizing ipv6 (as a general concept) should improve the user experience.
How long do you think it will take for the P2P software authors to transition over to IPv6? I'll bet that P2P users will be a lot more likely to use IPv6 over Aunt May checking her email once a day. Niall. -- The virus contained in this message was not detected.
On 15/10/2008, at 8:56 AM, Niall Donegan wrote:
Scott Doty wrote:
After all, if most p2p traffic is v4, prioritizing ipv6 (as a general concept) should improve the user experience.
How long do you think it will take for the P2P software authors to transition over to IPv6? I'll bet that P2P users will be a lot more likely to use IPv6 over Aunt May checking her email once a day.
Sorry, it happened already. Presentation I gave at APNIC26 on the subject: http://www.apnic.net/meetings/26/program/ipv6/ward-ipv6-stats.pdf A Teredo/6to4 relay in Finland operated by CSC/FUNET: http://www.braintrust.co.nz/temp/csc-funet-teredo-6to4.png Note the massive increase of traffic. It ramps up the *day* uTorrent 1.8 came out. I have long said: 1) IPv4 will continue to exist for web/email/etc. servers. 2) When ISPs run out of IPv4 space and start NATing their customers, IPv6 will become the new way to do applications that require end-to- end. This is *why* Teredo exists - so Microsoft applications that require end-to-end can get it, without having to implement a NAT traversal stack in each and every app. Teredo is now available on most OSes. 3) (2) is clearly happening already, with bit torrent applications. It's an easy way to do NAT traversal for free. 4) When IPv6 is widely enough supported, then maybe someone will run a web/email/etc. server on IPv6 only. There is a lot of work to attempt to keep IPv4 end-to-end alive when SP-NAT happens. Personally, I think attempting to prolong IPv4 end-to- end is a waste of time when IPv6 does it already, and all these proposals require applications to be updated to support dynamic ports and things. If you're updating the application, just make it support IPv6. For most applications this is trivial. -- Nathan Ward
If P2P became IPV6, and therefore universally endpoint addressable, and therefore seeded by every download, as opposed to solely seeded by those who have enough clue to configure the inbound ports through their IPV4 NAT, then the bandwidth problem should solve itself, at least for the widely popular downloads. Assuming every downloader could also seed, for the popular stuff, the traffic would be primarily local, which is not where the bottleneck is in most cases. Of course, this is assuming that the reason the SPs are opposed to P2P isn't related to their desire to derive revenue from the content that is non-ratable if delivered over P2P, but that is a different topic.
-----Original Message----- From: Nathan Ward [mailto:nanog@daork.net] Sent: Tuesday, October 14, 2008 3:33 PM To: Niall Donegan Cc: nanog@merit.edu Subject: Re: spurring transition to ipv6 -- make it faster
On 15/10/2008, at 8:56 AM, Niall Donegan wrote:
Scott Doty wrote:
After all, if most p2p traffic is v4, prioritizing ipv6 (as a general concept) should improve the user experience.
How long do you think it will take for the P2P software authors to transition over to IPv6? I'll bet that P2P users will be a lot more likely to use IPv6 over Aunt May checking her email once a day.
Sorry, it happened already.
Presentation I gave at APNIC26 on the subject: http://www.apnic.net/meetings/26/program/ipv6/ward-ipv6-stats.pdf
A Teredo/6to4 relay in Finland operated by CSC/FUNET: http://www.braintrust.co.nz/temp/csc-funet-teredo-6to4.png
Note the massive increase of traffic. It ramps up the *day* uTorrent 1.8 came out.
I have long said: 1) IPv4 will continue to exist for web/email/etc. servers. 2) When ISPs run out of IPv4 space and start NATing their customers, IPv6 will become the new way to do applications that require end-to- end. This is *why* Teredo exists - so Microsoft applications that require end-to-end can get it, without having to implement a NAT traversal stack in each and every app. Teredo is now available on most OSes. 3) (2) is clearly happening already, with bit torrent applications. It's an easy way to do NAT traversal for free. 4) When IPv6 is widely enough supported, then maybe someone will run a web/email/etc. server on IPv6 only.
There is a lot of work to attempt to keep IPv4 end-to-end alive when SP-NAT happens. Personally, I think attempting to prolong IPv4 end-to- end is a waste of time when IPv6 does it already, and all these proposals require applications to be updated to support dynamic ports and things. If you're updating the application, just make it support IPv6. For most applications this is trivial.
-- Nathan Ward
On 15/10/2008, at 6:19 AM, Scott Doty wrote:
Just wondering: what if we gave ipv6 traffic "mucho priority" over ipv4 traffic, then tell our user communities that ipv6 provides a better quality network experience, including (hopefully) faster page loads, & lower video game pings?
I think by the time we've put carrier NATs everywhere the users will notice that all by themselves, and we won't need to tell them anything. - mark -- Mark Newton Email: newton@internode.com.au (W) Network Engineer Email: newton@atdot.dotat.org (H) Internode Systems Pty Ltd Desk: +61-8-82282999 "Network Man" - Anagram of "Mark Newton" Mobile: +61-416-202-223
It's a good point that you brought up. Even though we already have IPv6 P2P (Nathan's post explains this in more detail), it would still be quite interesting to provide IPv6 as a higher class of traffic within service provider networks. Quite likely 6to4 relays and native IPv6 traffic is best effort today (ie. the same as IPv4 inet). I think operators would need to consider the financial implications of placing this traffic ahead of their current revenue generating services. Possibly instead of prioritizing the traffic, if the ISPs that normally police traffic from CE's could provide a higher policed rate for IPv6 traffic, so the experience is significantly different even in times where there is no congestion. Truman On 14/10/2008, at 3:49 PM, Scott Doty wrote:
We've had one presentation on the "unfairness" of p2p traffic, which (the presenter says) will eventually swamp us.
Then just now, we had the presentation & subsequent discussion re: ipv6 adoption.
Just wondering: what if we gave ipv6 traffic "mucho priority" over ipv4 traffic, then tell our user communities that ipv6 provides a better quality network experience, including (hopefully) faster page loads, & lower video game pings?
With such policies in place, folks wouldn't want to stay with the "old, slow" v4 traffic...and could be a significant selling point.
After all, if most p2p traffic is v4, prioritizing ipv6 (as a general concept) should improve the user experience.
Anyway, was just an idea, please pardon me if this has been discussed before, or sounds nutty...
Thanks,
-Scott
On 10/16/08, Truman Boyes <truman@suspicious.org> wrote:
It's a good point that you brought up.
Even though we already have IPv6 P2P (Nathan's post explains this in more detail), it would still be quite interesting to provide IPv6 as a higher class of traffic within service provider networks.
As long as none of your ipv6 traffic transits across anything from British Telecom as it is not supported on their 21st Century Network <http://aaisp.blogspot.com/2008/10/bts-21st-century-network-apparently-not.html> mike
As long as none of your ipv6 traffic transits across anything from British Telecom as it is not supported on their 21st Century Network
<http://aaisp.blogspot.com/2008/10/bts-21st-century-network-ap parently-not.html>
The distinction between supported, and unsupported is that when something is supported by a company it means that the company has explicitly agreed to provide the supported feature and is getting paid for it. This is marketing/sales lingo. Given that NANOG is a technical forum, I don't know why you bring this up. Clearly we do technically support IPv6 on 21CN because we have customers using it. That's how we discovered a Cisco bug that causes certain IPv6 packets to be mangled, and like all bugs that ISPs find in vendor equipment, it is reported and will eventually get fixed. This illustrates why it is important to deploy IPv6 now, even if it is only for technically clued-in customers who will participate in bug finding, etc. There are a lot of technologies involved here, and clearly we need to exercise some of the access technologies like L2TP a lot more to shake out the bugs and get them fixed. That isn't something that any one ISP can do on their own. It requires every ISP to take IPv6 seriously enough that they deploy it in their labs and offices, then report every bug and issue to vendors asking them to make sure these things are fixed before the end of 2010. This is how the IPv4 Internet managed to scale through a period of 1500% annual growth. It was the cooperation between ISPs, vendors and researchers, mediated by NANOG as a forum, that enabled the Internet to become as important as it is today. --Michael Dillon
On 18/10/2008, at 12:18 AM, Michael Simpson wrote:
On 10/16/08, Truman Boyes <truman@suspicious.org> wrote:
It's a good point that you brought up.
Even though we already have IPv6 P2P (Nathan's post explains this in more detail), it would still be quite interesting to provide IPv6 as a higher class of traffic within service provider networks.
As long as none of your ipv6 traffic transits across anything from British Telecom as it is not supported on their 21st Century Network
<http://aaisp.blogspot.com/2008/10/bts-21st-century-network-apparently-not.ht...
Yeah, except Teredo and 6to4 solve that problem for us today and are enabled by default on Vista and are used by p2p applications. Next. -- Nathan Ward
participants (9)
-
Mark Newton -
Michael Simpson -
michael.dillon＠bt.com -
Nathan Ward -
Niall Donegan -
Scott Doty -
Skywing -
Tomas L. Byrnes -
Truman Boyes