Routing Between Peers BGP

1. Routing Between Peers (BGP) Chapter 14

2. Routing Update Protocol Scope No routing protocol can scale to all routers in the Internet Internet routers cannot communicate directly Not all routers are managed by the same authority Therefore, a single routing protocol is impossible

3. Determining a Practical Limit on Group Size There are two issues: delay and overhead. Consider the maximum delay until all routers are informed about a change when they use the distance-vector protocol. N routers, N steps. Overhead. Meta information.

4. Determining a Practical Limit on Group Size (continued) Heuristic: it is safe to allow up to a dozen routers across a WAN and five times as many across a set of LANs This is general advice Manages use a traffic monitoring scheme Network utilization Overhead Too many routers?

5. Extra Hops An important lesson from the early Internet If a router outside the group (non-core) uses a member of the group as a default, routing will be suboptimal

6. Extra Hops (continued) http://www.calvin.edu/figure-14.1.pdf The extra hop problem A mechanism is needed to allow nonparticipating routers to learn routes from routers

7. Autonomous System Concept An authority guarantees that internal routes remain consistent and viable Chooses one of its routers to exchange information with the outside world

8. Exterior Gateway Protocols An autonomous system advertises network reachability to the outside EXPs are not really routing protocols BGP (Border Gateway Protocol) has evolved through four versions BGP = BGP-4

9. EGP (continued) Each autonomous system designates a router (gateway) near the edge that will speak BGP on its behalf http://www.calvin.edu/~lave/figure-14.2.pdf

10. BGP Characteristics Inter-Autonomous System communication Coordination among multiple BGP speakers (iBGP) to provide consistency

11. BGP Characteristics (continued) Propagation of Reachability Information Next-Hop Paradigm Why? Policy Support Reliable Transport using TCP

12. BGP Characteristics (continued) Path Information Avoids cycles Incremental Updates Support for Classless Addressing (CIDR)

13. BGP Characteristics (continued) Route Aggregation A single entry represents multiple, related destinations Authentication

14. BGP Funtionality and Message Types Three basic functions Peer acquisition and authentication (Won�t you be my neighbor?) Reachability (yes/no) Verification of connections

15. BGP Funtionality and Message Types (continued) http://www.calvin.edu/~lave/figure-14.3.pdf

16. BGP Message Header http://www.calvin.edu/~lave/figure-14.4.pdf Length: 19 through 4096 octets Marker is used for synchronization

17. The Marker BGP is using a stream protocol Boundaries Errors in length

18. BGP Open Message http://www.calvin.edu/~lave/figure-14.5.pdf HOLD TIME Maximum time the receiver should wait for a message. If exceeded, flush routes learned

19. BGP Open Message (continued) Parameters Authentication (type) Larger AS numbers KEEPALIVE functions as an ACK to an OPEN

20. BGP Update Message Advertise new destinations or withdraw previous advertisements http://www.calvin.edu/~lave/figure-14.6.pdf

21. Compressed Mask-Address Pairs Destinations are given in a list of IP addresses and require a corresponding list of masks http://www.calvin.edu/~lave/figure-14.7.pdf

22. Compressed Mask-Address Pairs (continued) LEN specifies the number of bits in the mask IP Address is also compressed according to the mask Eg, 153.106

23. BGP Path Attributes BGP is not a pure distance-vector protocol. Path attributes Factored How was the path information learned A list of ASes along the route

24. BGP Path Attributes (continued) Path attributes Allow receiver to check for forwarding loops. Receiver is in the list of intermediate ASes Policy constraints Source of all routes How do I know?

25. BGP Path Attributes (continued) Path attributes field (type, length, value) http://www.calvin.edu/~lave/figure-14.8.pdf Length field and its size

26. BGP KEEPALIVE Message Standard header, no data TCP does test a connection, it only reports a failure to deliver data Larger updates when necessary

27. Information From the Receiver�s Perspective Policies Routing Issue Must report routes that are optimal from the outsider�s perspective http://www.calvin.edu/~lave/figure-14.10.pdf

28. The Key Restriction An exterior gateway protocol does not communicate or interpret distance metrics even if they are available

29. The Key Restriction (continued) BGP cannot be used as a routing algorithm No way to compare Should only advertise routes that should be followed

30. The Key Restriction: Consequences BGP does not provide for the simultaneously use of multiple paths. All traffic goes one way BGP does not support load sharing

31. The Key Restriction: Consequences (continued) As a special case, BGP is inadequate for optimal routing in an architecture that has two or more WANs connected at multiple points

32. The Key Restriction: Consequences (continued) All ASes must agree on a consistent scheme for advertising reachability. BGP alone will not guarantee consistency.

33. The Internet Routing Architecture In the original Internet routing architecture, the core guaranteed consistent routing information because it had one route to any destination.

34. The Internet Routing Architecture (continued) IXPs are known as Network Access Points. The two ISPs enter into a peering agreement. A private peering represents the boundary.

35. The Internet Routing Architecture (continued) Upstream, downstream, or transit. ISPs use Routing Registries which maintain information about which ISPs own which address blocks.

36. The Internet Routing Architecture (continued) ISP A advertises reachability to N. But does A own N? No central authoritative registry. Black holes.

37. BGP Notification Message Used for control and errors http://www.calvin.edu/~lave/figure-14.11.pdf http://www.calvin.edu/figure-14.12.pdf http://www.calvin.edu/~lave/figure-14.13.pdf