Internet routing

Structurally, the Internet is a network of interconnected networks. Each network is called an autonomous system (AS). An AS is a network of routers managed by a single administrative entity. We will use the terms AS and Internet service provider (ISP) interchangeably although some large ISPs are divided into multiple ASes for ease of administration. The ASes may either peer privately or through a network access point (NAP) where several ASes interconnect at a common exchange point. ASes are organized as a tiered hierarchy. Tier-1 ASes are at the top level of this hierarchy and provide transit service typically to tier-2 ASes that in turn provide transit typically to tier-3 ASes and so on. As the bottom of this hierarchy are stub ASes that provide connectivity to end users. Tier-1 ASes are large and it is estimated that there are only about ten of them in the Internet today.The AS structure is not a strict hierarchy, i.e., there are interconnections between ASes at the same tier. A tier-3 AS may also obtain transit service directly from a tier-1 AS. In fact, the tier to which an AS belongs is often difficult to classify and tiers only provide a loose characterization of the AS hierarchy. A better characterization of the AS hierarchy is in terms of the economic relationship between interconnected ASes. An AS can be a provider, peer, or customer of another AS as determined by the economic relationship between them. A customer pays a provider for transit service. However, peer ASes typically do not charge each other for transiting traffic between their customers. The exact nature of the economic relationship between ASes is typically a trade secret and may evolve over time. Note that a customer always pays a provider irrespective of which direction (from/to the customer) the traffic flows. This aspect of Internet service pricing is fundamentally different from long-distance telephone services where the initiator of a call pays for the call while the recipient typically does not. The difference in the pricing schemes between Internet service and telephone service can be attributed to the nature of their traffic and resource allocation schemes. A telephone call has a call setup phase that clearly identifies the initiator and traffic flow in both directions is more or less balanced, so it suffices to charge just the initiator. Before caller identification became widespread, charging the initiator also avoided the scenario where a receiver of the call had to worry about the cost of the call without knowing its origin. In contrast, Internet traffic does not lend itself to an initiator-based pricing model. Consider a typical Web transaction where an end-user downloads content from a web site. One would be inclined to identify the end-user as the initiator of the connection, but most of the traffic flows from the web server to the end-user. Who should pay for this traffic and how does the network identify this entity? Who is responsible for additional TCP connections created by programs at either end? In the Internet, both parties—the end-user and the web server—pay their respective ISPs. The network layer is stateless and does not attempt to account for resource usage on a per-connection basis. Instead, economic settlement happens