Routing of Internet

Internet Service Providers connect customers (thought of at the "bottom" of the routing hierarchy) to customers of other ISPs. At the "top" of the routing hierarchy are ten or so Tier 1 networks, large telecommunication companies which exchange traffic directly "across" to all other Tier 1 networks via unpaid peering agreements. Tier 2 networks buy Internet transit from other ISP to reach at least some parties on the global Internet, though they may also engage in unpaid peering (especially for local partners of a similar size). ISPs can use a single "upstream" provider for connectivity, or use multihoming to provide protection from problems with individual links. Internet exchange points create physical connections between multiple ISPs, often hosted in buildings owned by independent third parties.[citation needed]
Computers and routers use routing tables to direct IP packets among locally connected machines. Tables can be constructed manually or automatically via DHCP for an individual computer or a routing protocol for routers themselves. In single-homed situations, a default route usually points "up" toward an ISP providing transit. Higher-level ISPs use the Border Gateway Protocol to sort out paths to any given range of IP addresses across the complex connections of the global Internet.[citation needed]
Academic institutions, large companies, governments, and other organizations can perform the same role as ISPs, engaging in peering and purchasing transit on behalf of their internal networks of individual computers. Research networks tend to interconnect into large subnetworks such as GEANT, GLORIAD, Internet2, and the UK's national research and education network, JANET. These in turn are built around smaller networks.[citation needed]
Not all computer networks are connected to the Internet. For example, some classified United States websites are only accessible from separate secure networks.