11.6 TRANSITION MECHANISMS

159

11.6TRANSITION MECHANISMS

It is clear that IPv6 is not going to supplant IPv4 in one hit on one ‘flag day’. So, if IPv6 is to be deployed in more than just isolated pockets, there must be mechanisms to operate both networks in parallel and to provide links between IPv6 ‘islands’ across the ‘sea’ of IPv4. There also must be a mechanism to provide connectivity between hosts using only IPv4 and those using only IPv6. Numerous mechanisms have been devised including dual-stack operation, Network Address Translation—Protocol Translation (NAT-PT) and IPv6 over various tunnelling mechanisms (e.g. IPv4, GRE, MPLS, etc.). Each has its benefits and disadvantages as described below.

11.6.1DUAL STACK

This mode of operation requires each interface to be configured with both an IPv4 and an IPv6 address. On this basis, each incoming packet is handled by the appropriate IP stack for the type of packet. In this mode, both networks are run in parallel. This is similar to running both IPX and IP on a single network. Neither protocol interferes with, or even recognizes the presence of the other.

While this does nothing for the conservation of IP addresses, this is the preferred method of running both IPv4 and IPv6 in a contiguous network and is almost certainly the mode in which most SPs will be operated. It provides native support to both protocols, thus IPv4 hosts use IPv4 to talk to other IPv4 hosts and similarly IPv6 hosts use IPv6 to talk to other IPv6 hosts. One downside is that this provides no inherent mechanism for translating between the two protocols. So IPv4 hosts cannot communicate with IPv6 hosts unless some external gateway mechanism between the two networks is provided.

11.6.2NETWORK ADDRESS TRANSLATION—PROTOCOL TRANSLATION

Having described the constraints of NAT, which have been used by many as a convincing argument for the development of IPv6, it seems somewhat ironic that one of the main mechanisms for joining the IPv4 and IPv6 worlds together is based on NAT-PT. NAT-PT is carried out by routers that support both IPv6 and IPv4. The fact that the hosts at each end of the conversation are not using the same version of IP is (or at least should be) totally opaque to the two hosts.

In addition to mapping the addresses in the headers of IP packets traversing the NATPT router, the router must also understand the protocols where IP addresses are carried within the data and, where appropriate, translate information within the data to match the mapping. For example, when a name server replies to a query from an IPv4-only host with an IPv6 address, the NAT-PT router must create the mapping between an IPv4 and the returned IPv6 address and then modify the DNS reply so that the IPv4-only host receives a DNS response containing IPv4 addresses.

160

IPv6

Finally, the mapping of IPv4 header information to IPv6 header information is not a simple task. It is not always possible to translate completely from one to the other.

The requirement for such interrogation of every packet traversing the NAT-PT router makes this process exceptionally CPU intensive. Any such process will inevitably limit the scaling of the network.

For all these reasons, it is better to use other transition mechanisms where such mechanisms are available.

11.6.3 TUNNELLING IPv6 IN IPv4

There are several mechanisms based on tunnelling of one sort or another. They include:

•configured IPv6 in IPv4 tunnels;

•automatic IPv6 in IPv4 tunnels;

•IPv6 in MPLS.

11.6.3.1 Configured IPv6 in IPv4 Tunnels

This was the first mechanism widely deployed for the connection of disparate IPv6 networks and formed the basis of the experimental 6bone. This mechanism requires the static configuration of tunnels at both ends. This is administratively time-consuming and also somewhat inefficient (a problem associated with any tunnelling due to the extra overhead of the encapsulating protocol).

11.6.3.2 Automatic IPv6 in IPv4 Tunnels

In order to overcome some of the administrative burden of configuring IPv6 in IPv4 tunnels, a scheme has been developed to use IPv4 compatible addresses to automatically tunnel IPv6 addresses through an IPv4 network. IPv4 compatible IPv6 addresses embed an IPv4 address into an IPv6 address. The last 32 bits of the address is actually represented as a standard dotted quad address of the form:

1234:5678:9ABC:DEF0:0000:0000:192.168.103.112

11.6.3.3 IPv6 in MPLS

Just as MPLS is capable of transporting IPv4 packets across networks not capable of forwarding them without the MPLS encapsulation, it is similarly capable of transporting IPv6 packets. The encapsulation of IPv6 packets is identical to that for IPv4 packets with the exception that there is a different label for IPv6 explicit null (2).