Importance of end-to-end IPv4 - research - please help
Hi all, I'm attempting to get a bead on the importance of end-to-end IPv4. By that I mean, home DSL user talking to another home DSL user. This is something that would break if we ran out of IPv4 space tomorrow, and had to start putting customers behind service provider NAT (SP-NAT). There's two ways I'm looking at doing this are: 1) Using a vendor box on loan to do p2p packet inspection for a month or so. This will tell us about how much "p2p[1]" traffic there is on a network, compared to non-p2p traffic. 2) Getting a packet capture from somewhere on a network for an hour, or whatever is feasible in terms of storage and processing power. The target of the capture would be traffic to/from a certain block of an ISPs end user type customers (so, a DSL pool probably). Analyse this and match it against dynamic address pools. - Anything going out to another dynamic pool (as determined by one of those dynamic pool lists) is something that would be broken by SP-NAT. - Any new incoming connections is something that would be broken by SP-NAT. If there's anyone that's interested in the following please let me know: a) Helping me with some research b) Getting some free intelligence on the type of traffic on your network (wave it in front of marketing, and drip feed them the pretty graphs whenever you want something from them) My intent is to publish the results stuff freely, publicly and widely. I'd even like to get to the point where we can do it regularly perhaps? Let me know if you're open to that. -- Nathan Ward [1] By this I mean file sharing, skype, etc. Stuff commonly identified with the "p2p" buzz word, as opposed to the technical peer-to-peer phrase.
SP-NAT won't necessarily break things. Many people sit behind NAT routers already. Many services work by relaying off a box that has a public IP address. So running a user through two layers of NAT isn't going to break that. Of course, it does mean they won't be able to setup port forwards to be able to run their own web server, p2p server, etc. But I guess that will become a "value add" service they can purchase ... -----Original Message----- From: Nathan Ward [mailto:nznog(a)daork.net] Sent: Tuesday, 5 August 2008 2:40 a.m. To: nznog Subject: [nznog] Importance of end-to-end IPv4 - research - please help Hi all, I'm attempting to get a bead on the importance of end-to-end IPv4. By that I mean, home DSL user talking to another home DSL user. This is something that would break if we ran out of IPv4 space tomorrow, and had to start putting customers behind service provider NAT (SP-NAT). There's two ways I'm looking at doing this are: 1) Using a vendor box on loan to do p2p packet inspection for a month or so. This will tell us about how much "p2p[1]" traffic there is on a network, compared to non-p2p traffic. 2) Getting a packet capture from somewhere on a network for an hour, or whatever is feasible in terms of storage and processing power. The target of the capture would be traffic to/from a certain block of an ISPs end user type customers (so, a DSL pool probably). Analyse this and match it against dynamic address pools. - Anything going out to another dynamic pool (as determined by one of those dynamic pool lists) is something that would be broken by SP-NAT. - Any new incoming connections is something that would be broken by SP-NAT. If there's anyone that's interested in the following please let me know: a) Helping me with some research b) Getting some free intelligence on the type of traffic on your network (wave it in front of marketing, and drip feed them the pretty graphs whenever you want something from them) My intent is to publish the results stuff freely, publicly and widely. I'd even like to get to the point where we can do it regularly perhaps? Let me know if you're open to that. -- Nathan Ward [1] By this I mean file sharing, skype, etc. Stuff commonly identified with the "p2p" buzz word, as opposed to the technical peer-to-peer phrase. _______________________________________________ NZNOG mailing list NZNOG(a)list.waikato.ac.nz http://list.waikato.ac.nz/mailman/listinfo/nznog
On 5/08/2008, at 2:44 AM, Philip D'Ath wrote:
SP-NAT won't necessarily break things.
Many people sit behind NAT routers already. Many services work by relaying off a box that has a public IP address. So running a user through two layers of NAT isn't going to break that.
Of course, it does mean they won't be able to setup port forwards to be able to run their own web server, p2p server, etc. But I guess that will become a "value add" service they can purchase ...
<insert religious debate and lots of corner cases here> There are /lots/ of things to consider, I'm attempting to get some real world numbers around at least some of them. Cheers, -- Nathan Ward
You may be interested in a study we did early this year with Alcatel. We looked at all the traffic to the DSL customers of a NZ ISP over three days and measured how many terminated incoming connections. The results were surprising (high). David Miles presented the results at Apricot specifically in the context of SP-NAT. I'm not at work and can't find the slides on the web which is why I'm not being specific about the results but if you contact Shane Alcock at Waikato he did the work and has the detailed results. Currently I'm trying to negotiate to capture traffic from another NZ ISP and if we are successful we would probably update this study and could look at other variations that you might suggest. Richard. Nathan Ward wrote:
Hi all,
I'm attempting to get a bead on the importance of end-to-end IPv4.
By that I mean, home DSL user talking to another home DSL user.
This is something that would break if we ran out of IPv4 space tomorrow, and had to start putting customers behind service provider NAT (SP-NAT).
There's two ways I'm looking at doing this are: 1) Using a vendor box on loan to do p2p packet inspection for a month or so. This will tell us about how much "p2p[1]" traffic there is on a network, compared to non-p2p traffic. 2) Getting a packet capture from somewhere on a network for an hour, or whatever is feasible in terms of storage and processing power. The target of the capture would be traffic to/from a certain block of an ISPs end user type customers (so, a DSL pool probably). Analyse this and match it against dynamic address pools. - Anything going out to another dynamic pool (as determined by one of those dynamic pool lists) is something that would be broken by SP-NAT. - Any new incoming connections is something that would be broken by SP-NAT.
If there's anyone that's interested in the following please let me know: a) Helping me with some research b) Getting some free intelligence on the type of traffic on your network (wave it in front of marketing, and drip feed them the pretty graphs whenever you want something from them)
My intent is to publish the results stuff freely, publicly and widely.
I'd even like to get to the point where we can do it regularly perhaps? Let me know if you're open to that.
-- Nathan Ward
[1] By this I mean file sharing, skype, etc. Stuff commonly identified with the "p2p" buzz word, as opposed to the technical peer-to-peer phrase. _______________________________________________ NZNOG mailing list NZNOG(a)list.waikato.ac.nz http://list.waikato.ac.nz/mailman/listinfo/nznog
One of the big talking points at last week's IETF was the variants on Carrier Grad NAT that are being touted. Opinions on this vary from "sublime" to "ridiculous". http://tools.ietf.org/id/draft-nishitani-cgn http://tools.ietf.org/id/shirasaki-isp-shared-addr APNIC has already said no to the proposal in the second draft to create a "new Net 10", and it's very controversial in the IETF. There's also the variant on this proposed as an IPv6 coexistence tool: http://tools.ietf.org/id/draft-durand-dual-stack-lite Although most p2p solutions and SIP-based protocols have ways to get round NAT, I think Nathan's suggestion is valuable. Facts are always helpful. Brian On 2008-08-05 02:39, Nathan Ward wrote:
Hi all,
I'm attempting to get a bead on the importance of end-to-end IPv4.
By that I mean, home DSL user talking to another home DSL user.
This is something that would break if we ran out of IPv4 space tomorrow, and had to start putting customers behind service provider NAT (SP-NAT).
There's two ways I'm looking at doing this are: 1) Using a vendor box on loan to do p2p packet inspection for a month or so. This will tell us about how much "p2p[1]" traffic there is on a network, compared to non-p2p traffic. 2) Getting a packet capture from somewhere on a network for an hour, or whatever is feasible in terms of storage and processing power. The target of the capture would be traffic to/from a certain block of an ISPs end user type customers (so, a DSL pool probably). Analyse this and match it against dynamic address pools. - Anything going out to another dynamic pool (as determined by one of those dynamic pool lists) is something that would be broken by SP-NAT. - Any new incoming connections is something that would be broken by SP-NAT.
If there's anyone that's interested in the following please let me know: a) Helping me with some research b) Getting some free intelligence on the type of traffic on your network (wave it in front of marketing, and drip feed them the pretty graphs whenever you want something from them)
My intent is to publish the results stuff freely, publicly and widely.
I'd even like to get to the point where we can do it regularly perhaps? Let me know if you're open to that.
-- Nathan Ward
[1] By this I mean file sharing, skype, etc. Stuff commonly identified with the "p2p" buzz word, as opposed to the technical peer-to-peer phrase. _______________________________________________ NZNOG mailing list NZNOG(a)list.waikato.ac.nz http://list.waikato.ac.nz/mailman/listinfo/nznog
participants (4)
-
Brian E Carpenter
-
Nathan Ward
-
Philip D'Ath
-
Richard Nelson