Autonomous Systems, IGP and EGP

 Interior Gateway Protocols IGP

Interior Gateway Protocols are the protocols that run inside an autonomous system. IGPs are all (or most) of the protocols you learned in school (or later in your career): RIP, OSPF, EIGRP, IS-IS,…

IGP can be redistributed into EGPs.

Exterior Gateway Protocols EGP

An exterior gateway protocol is used to route packets between autonomous systems. The most famous one is BGP.

EGPs can be redistributed into IGPs.

A definition of Autonomous Systems

An autonomous system (AS) is a group of IGPs that are administered by the same authority. We can’t really define autonomous systems as a number of routers running one IGP, because there are companies that merge, and their internetworks form one single autonomous system.

Autonomous system can mean also the routing process domain. For example, a router that is running an OSPF process number 1 is an autonomous system. Another OSPF process that runs on the same router is another autonomous system. Redistributing between -let’s say- BGP routes and OSPF routes is in fact redistributing routes between two autonomous systems.

Autonomous Systems in the real world

Usually a company is assigned one autonomous system. However, some big companies -such as Google- have more than one AS.

Transit AS and Transit path

An autonomous system is called a Transit AS if packets from one AS flow through it to reach another AS.

Take a look at the figure. We have three autonomous systems AS1, AS2 and AS3. If packets from AS1 transit through AS2 to reach AS3, then AS2 is a transit AS.


The path that is forged between the source AS and the destination AS across the transit AS is then called Transit path.


An ISP can provide transit network path for Internet traffic. However, it is not accepted for a enterprise client network to be a transit network.

Autonomous system number (ASN)

In the past, the autonomous system number ASN was written in two bytes. Nowadays, with the apparition of a new RFC, the ASN is written on four bytes.

Each byte ranges from 1 to 65535. So the biggest ASN value you can have is “65535.65535”.

The four-byte notation is backward-compatible with the two-byte notation. So if you have ASN 100, it can be written as 0.100.

Note that the range [64512 – 65535] is called private AS numbers. These private AS numbers are used to interconnect a company to its ISP and are not routed into the global Internet. In fact, when an ISP has multiple clients, it can assign to them unique private ASN. However, the ISP does not propagate the private ASN’s into the global Internet.

How to determine the Autonomous System

Given a public IP address, it is possible to determine the correspondent autonomous sytem. We do that online with a nice tool provided by Team Cymru.

  • We need to select the RIR that corresponds to your country. For example, if my country is Germany, I go to RIPE.



  • then, go to
  • select your RIR. Here it is RIPE NCC


  • All IP-to-autonomous system mapping are listed:


Tier 1 ISP, Regional ISP and Access ISP

There are three main categories of ISPs:

  1. Tier-1 ISP
  2. Regional ISP
  3. Access ISP
  • Tier 1 ISPs connect to each other in a full-meshed fashion and free-settlement, i.e. they do not charge fees for interlinking between each other and exchanging packets.
  • Regional ISPs are customers of Tier 1 ISPs. They pay Tier 1 ISPs to route packets across their links. Regional ISPs make peering links between each other that are usually free of charge. Besides, ISP autonomous systems are never transit AS for their peers.
  • Access ISPs are the ones you and me connect to. They are customers of Regional ISPs.
Figure: Interconnections between Tier1 ISPs, Regional ISPs and Accss ISPs

Customer, Provider and Peer

The relationship between these types of ISPs is a customer/provider/peer relationship; an ISP can take the role of a customer, a provider, both (customer and provider), and peer.

ISP_A is a customer of ISP_B when ISP_A purchases access to networks either managed by ISP_B or reachable via ISP_B; ISP_B is then the provider and AS_B is a transit AS. In this case, ISP_A has to purchase transit traffic bandwidth from ISP_B. Needless to say, most -if not all non-Tier-1 ISPs – need to purchase at least some transit traffic bandwidth.

ISP_C and ISP_D are peers when they can reach each other’s networks for free. However, if ISP_D were peering with another ISP_X, then ISP_C won’t see the networks of ISP_X through the peering link.

In terms of cost:

  • subscriber traffic is free of charge
  • Peer traffic is almost free of charge, if we neglect the cost of connecting to the peer AS (switch port and the interconnection link)
  • Transit traffic is the most expensive.
The customer/provider/peer relationship between ISPs ©




Be First to Comment

Leave a Reply

Your email address will not be published. Required fields are marked *