In our case, we have: > * many millions of edge devices (similar to your UE) > * thousands of routers > * thousands of servers
With those numbers in mind, it is easy to see that the pain is in finding IPv4 addresses for edge devices, not for routers or servers. In our case, taking IPv4 away on routers and servers would only saves a very small number of addresses compared to what is needed for the edge devices... Removing IPv4 from the routers may have the side benefits of simplifying network management, however one has to be very confident in the quality of the v6 equipments/management software before doing so... And if the majority of the traffic is still IPv4, one may question the validity of the approach. The path of least resistance seems to be, at least form my perspective, to get dual-stack to the network (already done by many operators), then get the servers ready for dual-stack operation so they can server any type of UE. On the UE front, you need dual-stack CAPABLE devices, that can be working either in a pure legacy IPv4 access network, in a complete dual-stack access network or in a v6-only access network, ie a dual-stack network where edge devices can only get a v6 address. In such scenario, with a dual-stack capable edge device configured only with an IPv6 address, as you have a v4 stack in your edge devices, the simplest thing to do is to use DS-lite to connect to the rest of the IPv4 Internet, and use IPv6 to connect to the IPv6 Internet. Communications stay in the same address familly and things are much simpler that way. Now, someone asked earlier why go through the pain of making the servers dual-stack if DS-lite actually solve the IPv4 address exhaustion issue? I previously answered in this list. This is because the more native end-to-end IPv6 traffic there is, the less burden on the DS-lite NAT infrastructure and other things app developers have to put in place to deal with the presence of NAT. - Alain.
_______________________________________________ Softwires mailing list [email protected] https://www.ietf.org/mailman/listinfo/softwires
