IP domain

Written on 5:10 PM by ooe

The principles of the IP domain are that an IP router in the user's domain (see Figure 5.16) communicates over the ADSL and ATM networks to the RAS. The RAS then tunnels the traffic over the IP network to a home gateway router on the service provider's premises. The home gateway router then presents an Ethernet interface to which the service provider interconnects. For higher speed aggregate access links, the home gateway router is on BT premises in a home gateway cluster, and the interface to service provider is provided by an NTE router on the service provider's premises.

The session from the user to the service provider is established when the user goes on-line by establishing a PPP session. The PPP packets will first be received at the RAS, which will check with the BT RADIUS that it is valid to forward this and to which service provider this should be sent. Once this is determined, the PPP packets are tunnelled to the service provider. The service provider needs to provide a RADIUS server such that they can perform additional security check on the user's request such as authentication. Depending upon the option selected by the service provider either the service provider's RADIUS or the home gateway will be responsible for allocating the IP address to the user for that session. Once the request has been validated and the IP address has been conveyed to the user, the PPP session is established with the home gateway router.

There is a limit on the number of simultaneous sessions that can be supported over the aggregate interface to the service provider. This limit ranges from 250 for smaller configurations up to a maximum of 8000.

The network has the ability to support either network address translation (NAT) or no-NAT options. NAT means that as packets pass through the router on the user's premises the address used is translated between that allocated for the session to the home gateway and that used for the PC. No-NAT means that router passes the address without translation.

All IP addresses are allocated for the session with the home gateway using the dynamic address allocation process. However, some service providers may support the ability to provide consistently served IP addresses. This option ensures the ‘dynamically allocated’ IP address is always the same. The option of consistently served IP addresses is required in combination with no-NAT in order to support some applications including H.323 Voice over IP (VoIP) applications.

There are a number of RASs located around the country, each serving a set of DSLAMs that in turn supports a set of users connected to the DSLAMs. The RAS that serves an exchange tends to be the one that is physically closest, though this does not have to be the case. As there is no control on the service providers that can be connected to and from any user, it is necessary that each RAS has the ability to establish tunnels to each service provider. These tunnels are established dynamically as a result of the first user request from that RAS to that service provider. Hence, when a user request is received and no tunnel exists to the appropriate service provider the PPP packets are held at the RAS until the tunnel to the service provider is established. The tunnel between the RAS and the home gateway router uses the L2TP protocol (layer 2 tunnelling protocol), though the service provider does not normally see this.

The normal termination of the L2TP tunnel is at the home gateway. However, there is also an option known as L2TP passthrough. This is an option where the L2TP tunnel does not terminate on the home gateway router, but is carried further into the service provider's network. This permits the service provider to control the user PPP sessions rather than having them controlled by the home gateway router. With this option there is no home gateway router provided on the service provider's premises. Instead the interface presented is a 155 Mbit/s STM-1 optical interface.