alex

Topic: IPv6 transition, really?

IPv6 seems to be near-and-dear to many folks' hearts, not so much for me. I don't see enough ISPs adopting IPv6 connectivity to their suite of offerings and it seems to just seem like a hobby for most people to play around with- latency is higher due to tunneling, 6to4, etc.

I find it interesting that according to Government guidelines, all agencies should be using IPv6 by now (well, as of June 30). According to various sources, this "deadline" was not reached on-time. I would assume this is because of inadequate IPv6 routing globally.

Some reads regarding the government move to IPv6:

http://www.whitehouse.gov/omb/egov/b-1- … .html#IPV6
http://www.gossamer-threads.com/lists/n … ers/103768

Will the world one day accept IPv6? If so, when do you think it will realistically become used?

Alexander McMillen - <amcmillen(at)sliqua.com>
Co-Founder - Oceanius Networks
President - Sliqua Enterprise Hosting
alex's Website

ftp

Re: IPv6 transition, really?

http://www.youtube.com/watch?v=_y36fG2Oba0

Case Closed!

01101110 01100101 01110110 01100101 01110010 01100111 01101001 01110110 01100101 01110101 01110000

alex

Re: IPv6 transition, really?

ftp wrote:

http://www.youtube.com/watch?v=_y36fG2Oba0

Case Closed!

Amen brother! cool

Alexander McMillen - <amcmillen(at)sliqua.com>
Co-Founder - Oceanius Networks
President - Sliqua Enterprise Hosting
alex's Website

martijn

Re: IPv6 transition, really?

Latency when tunneling <-- uhh tunneling is only there to aid in a smooth transition. You do not want tunneling when its deployed. My provider gives me a tunnel, but I cant wait until they deploy it native, I have the equipment for it.

Putting the fun back in funnel.
martijn's Website

anotherjason

Re: IPv6 transition, really?

There was an interesting talk on IPv6 at HostingCon last week by one of the guys from ARIN.  The gist of his talk was "no really, we are running out - it's real this time."  He said that there are currently 39 /8's that haven't been allocated worldwide, 12 were used last year and they're on track to give out 14 this year.  In particular use in India and China are skyrocketing, so they expect these rates to continue to increase which only gives a couple of years before the v4 space is completely allocated.  Now, how much of those blocks are actually *in use* is unclear, but apparently they've been getting stricter about their allocation policies so I'd imagine that most will be in use pretty soon.

The adoption rate seems to be really low - he said most of the major backbone providers are on board but not that much has been done beyond that.  Seems like the major driver to adoption will be as broadband providers have to start putting up v6-only networks - again, this will probably happen in India and China first due to their growth.  That will likely drive content providers to start support v6 so that their content can be accessed by all those new folks.

Definitely seems like it's going to be a painful transition to go through tho.  This might be a gift to MS tho in terms of forcing Vista adoption.

alex

Re: IPv6 transition, really?

Latency for me going from home -> tunnel -> sliqua isn't too bad, but then again, the tunnel itself and the router it's going to are basically in the same building... going elsewhere on the internet slows down things substantially due to lack of IPv6 transit and peering across the globe.

Going to the Sliqua network via IPv6 tunnel:

traceroute6 to 2001:4830:1600:144::2 (2001:4830:1600:144::2) from 2001:470:8:9a:21b:63ff:febc:6b3, 30 hops max, 12 byte packets
1  2001:470:8:9a:21b:63ff:fef2:239b  0.711 ms  0.363 ms  0.296 ms
2  sliqua-1.tunnel.tserv13.ash1.ipv6.he.net  10.73 ms  8.48 ms  10.371 ms
3  2001:470:0:90::1  9.694 ms  9.897 ms  9.715 ms
4  ibr01-ve96.asbn01.occaid.net  8.174 ms  10.004 ms  10.684 ms
5  equi6ix.dc.hotnic.net  10.613 ms  10.56 ms  10.452 ms
6  sixxs-asbnva-gw.customer.occaid.net  10.496 ms  13.254 ms  11.355 ms
7  cl-325.qas-01.us.sixxs.net  14.327 ms  12.307 ms  11.604 ms

vs. IPv4

fireside:~ sliqua$ traceroute fw1.res1.sliqua.com
traceroute to fw1.res1.sliqua.com (67.217.162.38), 64 hops max, 40 byte packets
1  172.16.1.1 (172.16.1.1)  0.629 ms  0.251 ms  0.196 ms
2  10.5.168.1 (10.5.168.1)  8.351 ms  6.488 ms  5.973 ms
3  ip72-219-223-241.dc.dc.cox.net (72.219.223.241)  10.863 ms  7.382 ms  7.604 ms
4  mrfddsrj01-ge110.rd.dc.cox.net (68.100.0.161)  15.416 ms  9.183 ms  8.016 ms
5  ashbbrj02-as0.0.r2.as.cox.net (68.1.1.232)  8.602 ms  11.975 ms  12.999 ms
6  border8.te2-1.bbnet2.wdc002.pnap.net (216.52.127.88)  10.328 ms  10.199 ms  10.096 ms
7  INTERNAP-10GE-1.COLO2.ASH1.NET.PTPTECH.NET (67.217.162.41)  19.947 ms  10.333 ms  9.657 ms
8  FW1.RES1.SLIQUA.COM (67.217.162.38)  9.761 ms  11.902 ms  10.184 ms

Not much of a difference with the above latency, etc.

However, to give a small example of latency increases over data center connections:

fw1:~# traceroute 216.218.252.170
traceroute to 216.218.252.170 (216.218.252.170), 30 hops max, 40 byte packets
1  VLAN243.COLO2.ASH1.NET.PTPTECH.NET (67.217.162.37)  0.942 ms  0.992 ms  1.047 ms
2  border8.tge2-4.pulltheplug-1.wdc002.pnap.net (216.52.125.25)  0.733 ms  0.794 ms  0.849 ms
3  core2.te5-2-bbnet2.wdc002.pnap.net (216.52.127.72)  0.602 ms  0.645 ms core2.te5-1-bbnet1.wdc002.pnap.net (216.52.127.8)  0.686 ms
4  cr1.wdc005.inappnet-62.core2.wdc002.internap.net (66.79.151.129)  0.631 ms  0.687 ms  0.744 ms
5  ash-ix.he.net (206.223.115.37)  1.036 ms  1.030 ms  1.101 ms
6  10gigabitethernet1-2.core1.nyc4.he.net (72.52.92.85)  6.572 ms  6.546 ms  6.603 ms

HOP 7 didn't respond over IPv4, but you can get an idea on the latency difference when compared to:

fw1:~# traceroute6 2001:470:0:12::1
traceroute to 2001:470:0:12::1 (2001:470:0:12::1), 30 hops max, 40 byte packets
1  gw-325.qas-01.us.sixxs.net (2001:4830:1600:144::1)  2.839 ms  2.805 ms  2.807 ms
2  sixxs-gw.hotnic.us.occaid.net (2001:4830:e6:7::1)  3.766 ms  3.751 ms  3.749 ms
3  core1.ash1.he.net (2001:504:0:2::6939:1)  4.345 ms  4.341 ms  4.344 ms
4  10gigabitethernet1-2.core1.nyc4.he.net (2001:470:0:36::2)  10.897 ms  10.882 ms  11.127 ms
5  loopback-1.core1.nyc1.he.net (2001:470:0:12::1)  11.687 ms  11.688 ms  11.674 ms
fw1:~#

Needless to say, if there was more transit/peering agreements in place for IPv6, the latency would be very similar to IPv4 - but I don't see enough people adopting it at this point. I'm sure most people can't tell a difference, but I have OCD about this stuff.. sad

Another thing that really cracks me up is that Vyatta hasn't adopted any IPv6 support on their router solution.. I was able to set up an IPv6 tunnel interface on the system outside of the CLI, but I can't actually route any space. Kinda sucks at this point.

Last edited by alex (2008-08-02 11:24:13)

Alexander McMillen - <amcmillen(at)sliqua.com>
Co-Founder - Oceanius Networks
President - Sliqua Enterprise Hosting
alex's Website

martijn

Re: IPv6 transition, really?

Well, as you so cutely accentuate in your post, you're most negative point is the latency of a tunnel. If you would tunnel IPv4 to something else it would introduce a similar latency. IPv6 is there to be used without a tunnel in the future, in order to get us there we are using tunnels to ease the transition. The more people have a negative opinion about a tunnel the less we are making progress on getting native IPv6. Really people should be more encouraged to make the transition to IPv6.

Putting the fun back in funnel.
martijn's Website

alex

Re: IPv6 transition, really?

anotherjason wrote:

The adoption rate seems to be really low - he said most of the major backbone providers are on board but not that much has been done beyond that.  Seems like the major driver to adoption will be as broadband providers have to start putting up v6-only networks - again, this will probably happen in India and China first due to their growth.  That will likely drive content providers to start support v6 so that their content can be accessed by all those new folks.

ARIN really needs to get strict about reclaiming under-utilized address space from the government, large corporations, and universities that were early to adopt IPv4. ARIN is getting really strict about offering out new allocations on IPv4 and will give you IPv6 all day long, however, there are so many IPv4 /8s that aren't being used efficiently (or at all) but are allocated out.

Examples of this include Fortune 500 companies giving all their workstations globally routable IPv4 addresses and filtering traffic going to them out on their routers/firewalls - when they could easily use NAT address space. I've seen a few large corporations and universities with /8s that I'm sure they'd never use in a million years, and to top that off they're wasting address space on other (smaller) allocations for their websites, etc.

With regards to countries like China adopting IPv6, I think that when this happens we'll have plenty of IPv4 space for the rest of the world and China will be stuck using 6to4. cool Then what reasons are there to adopt IPv6?

Alexander McMillen - <amcmillen(at)sliqua.com>
Co-Founder - Oceanius Networks
President - Sliqua Enterprise Hosting
alex's Website

anotherjason

Re: IPv6 transition, really?

alex wrote:

ARIN really needs to get strict about reclaiming under-utilized address space from the government, large corporations, and universities that were early to adopt IPv4. ARIN is getting really strict about offering out new allocations on IPv4 and will give you IPv6 all day long, however, there are so many IPv4 /8s that aren't being used efficiently (or at all) but are allocated out.

Agreed.  Part of the problem is that ARIN doesn't really have any power to force things to happen - they're working with organizations that have received large grants, but don't have any authority to force anyone to give up addresses that they're not using.  I think they said they had reclaimed a few /8's in the past year, but it's definitely not enough to keep up with growth.