BGP (Border Gateway Protocol) Summary

(e)BGP is a type of EGP (Exterior Gateway Protocol) - in fact the only one in use today and is used to provide roouting information accross the internet (across AS's.) opposed to an IGP (Interior Gateway Protocol) such as EIGRP or OSPF that provides routing information accross nodes in a single AS (Aoutonomous System.)

One of the fundamental differences between BGP and other routing procotols such as EIGRP, OSP etc. is that both parties (routers) must explicitly define a membership with each other.

There is also another type of BGP called iBGP - that as the name suggests is used

iBGP and eBGP both make use of AS (Autonomous System) numbers. Currently AS numbers are defined a 16-bit hence allowing a maximum of 65535 ASN's - although there are proposals for this to be raised to 16bit: https://tools.ietf.org/html/rfc6793

AS Numbers 1 - 64,495 are used for public use (eBGP) and 64,512 to 65,534 are reserved for private use (iBGP.)

The vast majority of publically (eBGP) available AS's are assigned to either ISPs and large enterprises. Although private AS's (iBGP) is typically applied within large ISP networks.

Typically we are used to routing protocols focussing on finding the optimal path to all destinations - although BGP differs somewhat as peering agreements between different ISPs can be very complex and as a result BGP carries a large number of attributes (metrics) with each IP prefix.

Some of the more commonly used are:

AS Path: The complete path outlining exactly which autonomous systems a packet would have to traverse to get to its destination.

Local Preference: Used in iBGP - if in the event there are multiple paths from one AS to another this attribute defines the preferred path.

Communities: This attribute is appended to a route being advertised to a nieghbor BGP router that provides specific instructions such as:

No-Advertise: Don't advertise the prefix to any BGP neighbors
No-Export: Don't advertise the prefix to any eBGP neighbors

BGP was built for security and scale-ability - although is certainly not the fastest routing protocol when dealing with convergence - hence internal routing is performed by a protocol with faster convergence such as OSPF and then external / internet routes are exchanged by BGP.

The diagram above shows a good (yet simplistic) representation of a service provider - we are firstly assigned our public ASN - we have a core network running iBGP between our core routers. We are also peering with two ISPs (i.e. we form an adjacency with their routers over eGBP) and also have two customers who use eBGP to peer with our network.

As a good introduction to BGP I would also highly reccomend taking a look at an article by the Internet Society.