An important motivation for the design of IPv6 was to reduce the global routing table size. The aggregation of IGP and EGP IPv6 routes should be better by design, but policies and network engineering decisions over the lifetime of IPv6 might show differently.
In the early 1990s, the global routing table size was seen as a major obstacle to the growth of the Internet, given that core routers might not have the horsepower and memory to handle the global routing table. Recent studies [RFC3221] have shown that the most important obstacle is not the size of the global routing table but its relative instability, caused by the visibility of the site routes in the global routing table. This visibility of site prefixes is used for traffic engineering and for site multihoming, which are two useful features for large networks.
To achieve site multihoming in IPv4, the site prefix is advertised by all the providers of the site. Figure 9.6 shows a site with the 192.0.2.0/24 prefix announced by its two providers to the Internet core.
Table 9.8 lists the entries in the global routing table for this site example.
In the global routing table, the site prefix has two specific entries, one for each provider. This enables the site to be fully redundant if one provider is not reachable, and also permits some load balancing. The example in Figure 9.6 uses portable addresses also called provider independent (PI) addresses, where the site owns the address space.