This is the second part of a two-part blog post. In part 1 we introduced Autonomous Systems and talked about BGP. In this part we cover service provider interconnection and how traffic traverses the globe across multiple service providers to reach its destination. You can view part 1 here.
BGP alone cannot deliver traffic originating from one AS to its destination. This is where network operators and inter-AS relationships come in. Network operators across the world, while competing, also cooperate to make global Internet routing successful for their customers. This interconnection allows for traffic originating from a source in one AS to traverse multiple AS’s that lie in between and reach its destination located in another AS.
Network operators can be broadly classified as Tier-1, Tier-2 or Tier-3. A Tier-3 operator is one that usually operates in a small geolocation and caters only to the local end customers. Tier-2 operators are more geographically spread and could operate across a state, region or country.
The Tier-1 operators cover almost the entire Internet and provide connectivity to regional and even countrywide providers, including ISPs, network service providers, cable companies, content delivery networks (CDN) and so on. The lower tier operators are customers of the Tier-1 operators and depend on them for interconnectivity. Interconnections between operators can be over direct or indirect connections and use either “peering” or “transit” to exchange traffic.
Peering can be established as paid or settlement-free (no costs involved) where direct exchange of traffic occurs between two network operators (or Autonomous Systems). Tier-1 providers peer with other Tier-1 providers to carry traffic from the other’s client networks.
When a network operator (or Autonomous System) carries traffic from another operator/AS that is destined for yet another AS, the interconnection is referred to as transit. The sender pays the transit network operator a transit fee. If a lower tier operator directly connects to a Tier-1 operator, they pay the Tier-1 operator an interconnection fee for carrying the traffic.
You can find more in-depth information about Internet Routing from this MIT study.
With the commercialization of the Internet came the requirement for commercial service providers to exchange traffic instead of depending on the publicly funded NSFNET, which was primarily meant for research and education. To address this need, the U.S. government commissioned four Network Access Points (NAPs).
The NAPs were designed to be a public network exchange facility where commercial service providers could interconnect and exchange traffic, usually over peering arrangements. Over time, as the NAPs became congested, service providers started shifting to privately operated exchange points – known as Internet eXchange points (IXPs) – for interconnection.
Similar to NAPs, IXPs enable interconnection between Autonomous Systems for Internet traffic exchange. By leveraging an IXP, many network operators can directly connect with an operator instead of paying for transit traffic to another provider, saving money. Today there are more than 400 IXPs in the world where AS-interconnection takes place.
In summary, end users send their traffic to a regional service provider, which can be a Tier-3 or Tier-2 operator. These lower tier operators in turn directly connect to a Tier-1 provider at an IXP or indirectly by sending traffic through a transit provider that has a direct connection with a Tier-1 provider. The Tier-1 provider then peers with another Tier-1 provider that can reach the destination for the Internet traffic. All of this inter-networking, or to be exact, inter-AS routing, is done using the Border Gateway Protocol (BGP).
And with that we complete the basics of Internet routing.