Interdomain Internet Routing

Internet infrastructure constructed by cooperation between ISPs with diverse internal structure and scope called Tier3 (local scope), Teir2 (regional scope) or Teir1 (global scope). Internet involves number of different types of autonomous systems AS that exchange route information using routing protocols BGP (Border Gateway Protocol, while IGP protocol can be used within each As such as RIP, OSPF, IS-IS and E-IGRP.

Forms of AS-AS interconnection
1) provider-customer transit is mostly
inter-AS relationship where the provider provides access to all destination in its routing table with charges for forwarding packets
ISP charges customers depending on the data rate in two ways; fixed pricing and changeable price according to the usage of the bandwidth

2) peering is typically for inter ISP relationship where the providers provide an access to a subset of each other’s routing tables (own transit customers and internal ISP address). This kind on interconnection is charge free and often established between competitors and it is emerged due to:
Number of Teir-1 ISP has been emerged with peering relationship with each other to avoid the control of singleTier-1 on all Internet traffic (in case if we have only provider-customer transit relationships). Such that the ASes related to different ISP don't have to pay money for any forwarding of packets between their customers. Also, peering provides kind of direct path between ASes that is necessary to achieve better end-to-end performance.

As a result, advertising a route by an ISP to its neighbor ISP means it agrees to forward packets to this route, thus ISP must take care to not advertise a route that impose it to act as a transit for packets

BGP was created to replace EGP routing protocol to provide decentralized which allowed the Internet to become a decentralized system. It supports Classless Inter-Domain Routing and uses route symmetrization to decrease the size of routing table. It's design goals are
· Scalability: BGP must ensure that the route advertisement scales well while parts of network going up and coming down
· It must provide to reasonable degree loop free paths
· It must allow ASes to apply their policy to make local decision like filtering and ranking or routes. At the same time keep AS's policy non disclosure
· BGP has been designed to provide routing between different administratively system

Routing Policy
1) Filtering to save or to make money by ISP
· The most important for ISP to advertise routes to its transit customers to as many other connected ASes as possible.
· If ISP received an advertisement from a peering ISP of a route to its transit customer, it will not advertise this route to other peering ISPs, because that may cause a peering ISP to use the advertising ISP to reach that destination which would expend ISP resources but not lead to revenue.

2) Ranking to import routes in routing table
There are many attributes that can be considered to import routes in routing table like
· Local Free attributes implies that the priority is for the customer route (since it has direct path to the customer), then peer route and finally provider route.
· Length of As attribute, actually I didn't figure out its meaning. I thought it means number of hops but I found out it's not
· MED is used when two ASes connected at many locations
· Routes learned via eBGP is preferred on that learned via iBGP

BGP is unique in using TCP as its transport protocol. BGP peers are established by manual configuration between routers to create a TCP session on port 179:
1. Establish a TCP session
2. Routers send OPEN message
3. Routers exchanges their tables (of course after applying ranking and filtering)
4. Then routers send
· UPDATE message (contains only the changes in its table since last update) to add or withdraw a route
· Or KEEPALIVE message periodically to maintain the connection (every 60 seconds by default). In case of absence this message the session terminated during hold time

When BGP is running within an (AS), it is referred to iBGP (Interior BGP) to exchange information about external routes, and it is called eBGP (Exterior BGP) when it is running between ASes. Routers on the boundary of one AS, exchanging information with another AS, are called border or edge routers. In the Cisco operating system, iBGP routes have an administrative distance of 200, which is less preferred than either external BGP or any interior routing protocol. Other router implementations also prefer eBGP to EGP, and IGPs to iBGP.

The paper focuses on interested subject and I wish if it discussed deeply