Hi, dear mailing list. On behalf of a medium retail ETTH ISP here in central Russia, I'd like to share a piece of CGNAT44 expirience with you. We are using it right from the start of the business for the 95% of our subscribers on their way to Internet. We've tried out several application approaches on this matter: single public IP with round robin port mapping to a 10.a.b/24 full of subscribers; random IP with session long port-block allocation of various ranges (64-1024 portblocks allocated per user at a time), while maintaining NAT behaviour requirements for TCP/UDP as per RFCs (major NAT traversal tricks as well). Number of ports required for a single user may vary in scale dramatically - simple websurfing takes 50-100 ports, while heavy p2p application can easily take 10K active connections, this Optimal port utilisation may be reached with up to 350-400 active users per public IP address without significant service degradation. This gives us around 2.5-3.5M IPv4 addresses required for a 1B users surfing internet and doing some serious p2p... So... an ISP can survive with this shortage and provide a "static public IP address" as a VAS - as it seems to me =) What bothers us is that it's still quiet difficult to get even as little as /21 IPv4 address as an additional allocation in case if we want to launch a new regional network... By the way - just while this discussion lasts, I've recieved 2 emails asking if we want to buy or sell any internet numbers =) 05.06.2012 2:22 пользователь "Lu Heng" <<a href="mailto:h.lu@anytimechinese.com" target="_blank">h.lu@anytimechinese.com</a>> написал: > > Hi Colleuages: > > Today I come across an very interesting article which I'd like to > share with your guys and see what your guys think about it. > > <a href="http://ilia.ws/archives/236-ISP-Popularity-by-Domain-Count.html" target="_blank">http://ilia.ws/archives/236-ISP-Popularity-by-Domain-Count.html</a> > > If what said in the article was true(which it should be as all the > data from the article should be public data), if all the cable network > start to use NAT, just in sense of domains, we might have IP supply > for few more decades. > > And it would be very interesting if someone from Ripe NCC can share > with us what is the most IP consuming business in the planet. > > In which, is that service can use NAT in sometime future? > > I talked to an UK telecom provider one day in a ripe lunch, he told me > that their network is already partly using NAT(sorry I didn't remember > his name, but I am sure he is on this mailing list), and he only need > a /21 for entire network. > > And another thing was, I heard from one of my friend in the Apanic > meeting, that someone is selling entire A class there for 10 USD/ IP. > > We don't know if there is already some successful story there > regarding IP sales. But seems to me, based on what happened on Asian > now, at least in China, as I heard from many of my colleagues there, > there was no real shortage there at this time. > > No body goes bankrupt because of no IPs left. > > So that raise an interesting point, since 60% of the world domain is > in fact wasted, is that the same story with IP usage. After all, we > allocated almost half of the pool before RIR even exists. > > The current way of IP distributing results a very noneffective way of > "past business" IP usage as well(e.g. someone changed their business > from cable business to an enterprise business, of course that guy will > not return Ripe NCC his additional free 2 M IPs). > > So when their is a real market for IPv4, and all the latency space > come to sale, will we last much much longer than everybody expected? > > We our-self have few dozen of enterprise customers, and they are > already paying 3-5 USD/ month /IP for years, as asking them to > re-program their software and re-provision their business into IPv6, > the cost will be enormous for them, and one more thing is, most > software writers don't really like IPv6, as it is hard to remember and > hard to type(it will be a whole lot easier if you just remember the IP > and type it every time you have to do so, rather than copy paste, we > all know the reason). So, even IP price eventually raise to 30USD even > 50USD, it is still very hard for them to switch it over to IPv6 as > long as they don't have IPv6-only client, because they are paying this > amount of money for IP per year anyway. > > Another thing is, one thing aside from RIR meetings and ISP meetings, > we didn't really hear a lot about IPv6. IPv6 come into developer is > still something new and interesting, and one guy I talked in the IPv6 > workshop in the Ripe64, a software developer for an Austria local > company, who don't even know that IPv6 has been around for almost a > decades. > > Since Ripe is almost finish it's public pools, my last question is, > will one day all the wasted IP address being effectively picked up > because of existing marketplace, and that market place will last us > another 3 decades before we really going to IPv6 ear? > > Go Ipv6 > Pro > Last forever(this reason doesn't really come into play because it this > can be a reason convince enterprise customer then it should be done 10 > years ago.). > unlimited amount of address space.(same as above) > cost reasons. > con: > need new router new config new practice and testing almost every part > of the business(from software to hardware). > very hard to remember and write. > (welcome to add more on this list) > > IPv4 > pro > you know how it works > it works for all of our business life. > it is easy to remember > con > cost reasons. > > But let's look at how much we are paying Ripe NCC now, for large ones, > they are paying more coffee in the office than they are paying Ripe > thing. So does that really hurts them when they pay 20USD per IP, look > at their margin and their current costs structure, I would say for > most business, it should be fine. > > If there is enough supply in the market for next decades, and keep the > price well below 50 USD per IP, I believe 99% of business would accept > this price and go on with their life. > > History already tell us most of us don't look too far to the > future(otherwise we are already there). > > Hope my a bit of 2 cents can get more interesting thoughts come around. > > -- > -- > Kind regards. > Lu > > This transmission is intended solely for the addressee(s) shown above. > It may contain information that is privileged, confidential or > otherwise protected from disclosure. Any review, dissemination or use > of this transmission or its contents by persons other than the > intended addressee(s) is strictly prohibited. If you have received > this transmission in error, please notify this office immediately and > e-mail the original at the sender's address above by replying to this > message and including the text of the transmission received. > > ---- > If you don't want to receive emails from the RIPE NCC members-discuss > mailing list, please log in to your LIR Portal account and go to the general page: > <a href="https://lirportal.ripe.net/general/view" target="_blank">https://lirportal.ripe.net/general/view</a> > > Click on "Edit my LIR details", under "Subscribed Mailing Lists". From here, you can add or remove addresses.