Sie sind auf Seite 1von 2

You need to understand that what follows is a gross simplification of BGP so you have to take it

that way.

BGP is actually very similar to RIP. In its core, just like RIP, it learns routes from different
neighbors (in BGP called peers), adds its own routes, decides what path toward a particular
destination is the best one, and for each known destination, sends this single best route to its
peers.

Principial differences between BGP and internal routing protocols are (this list is not
comprehensive):

1. BGP uses TCP as the transport protocol, and consequently, it uses only unicast
communication. Internal routing protocols (EIGRP, OSPF, RIP) use UDP or plain IP, and
they use multicasts to talk to multiple directly connected neighbors at once.
2. In BGP, every peer's address has to be specified manually. BGP does not discover its peers
dynamically. A BGP session can only be created between two routers if they are both
configured for mutual peering, referencing the other's address.
3. As a consequence of the previous properties, BGP peers can be multiple hops apart. They
do not need to be directly connected - as a matter of fact, very often, they aren't.
4. BGP, as opposed to any internal routing protocol, has a concept of an autonomous system -
a method of administratively defining which part of a larger network is under my control (my
autonomous system) and which parts are under someone else's control (their autonomous
systems).
5. BGP operation differs depending on whether a router talks to a peer in the same
autonomous system, or in a different autonomous system. One of the most important
differences is the advertising of known routes: If talking to a peer in a different autonomous
system (also called an external peer), the router can advertise any destination it knows
about. If, however, talking to a peer in the same autonomous system (also called an internal
peer), only routes learned from external peers and own routes can be advertised - in other
words, a route learned from an internal peer cannot be advertised to another internal peer.
This is a method of avoiding internal routing loops.
6. BGP does not have a simple metric to decide whch path is the best. Instead, it advertises an
extensive set of attributes with each route and uses a complex algorithm consisting of up to
13 steps to decide which path is the best.
7. BGP messages are Open, Keepalive, Update, Notification, and Route-Refresh. The Open
message is sent only at the beginning of the BGP session over an already established TCP
connection. This Open message allows routers to compare their settings, decide what kinds
of routes (IPv4, IPv6, etc.) shall be carried over this session, and negotiate their support of
various BGP capabilities. The Keepalive message is sent periodically to verify whether the
TCP connection is still up and running. The Update message is used to advertise and
withdraw routing information. Finally, the Route-Refresh message asks the peer to send all
its routes of a particular type again.
8. It is difficult to talk about BGP tables because different sources claim that BGP maintains
different sets of tables. However, what you are going to see on Cisco routers is the table of
neighbors, the BGP table (also called BGP Routing Information Base, or BGP RIB - this one
is similar to EIGRP's topology table), and then, of course, the router's normal routing table
that receives the best paths determined by BGP's best path selection algorithm.

Das könnte Ihnen auch gefallen