6 - BGP Flashcards
Why use EGP to link ASes?
Scaling to large networks (hierarchy, limit scope of failure)
Define admin boundaries (hide info, make own routing choices)
Policy (Control network reachability)
Reasons not to use IGP to link ASes
If the other ISP has routing problems:
- Instability will ripple (backbones)
- Leaked prefixes leak into your backbone
- Won’t scale
Hard to filter routes.
BGP
Border Gateway Protocol
BGP focuses on…
Routing policy, not topology
Binds network groups
BGP pros
Good route filtering capabilities
Ability to isolate from other’s problems
BGP can be used to keep traffic local. Why is this useful?
Think about cost
Traffic doesn’t have to go over your expensive line but is instead peered locally
BGP Policy
Control how you accept and send routing updates to neighbours
Peering
COnnecting to other local ISPs directly.
Transit
Connecting through ISPs to other ISPs.
At a cost
Why Peer?
Reduce upstream transit costs
Local Traffic stays local
Increase International bandwidth
Can increase performance
May be the only way to connect customers to some part of the internet (tier 1)
Why not Peer?
- Peers are usually competition
- Requires periodic renegotiation
What provides internal connectivity in an AS?
IGP
BGP OPEN
Opens TCP conn. to peer and authenticates
BGP UPDATE
“Announcement”: prefix is reachable
“Withdraw”: prefix is not reachable
BGP KEEPALIVE
keeps connection alive in absence of UPDATES
serves as ACK to OPEN request
BGP NOTIFICATION
reports errors in previous msg
closes connection
BGP basics
2 items. Updates and Protocol
Incremental Updates
- Send one copy of RIB then changes
Path Vector protocol
- Keeps track of the AS path of routing information
neighbor
BGP peer
NLRI
NLRI - Network Layer Reachability Info
Router ID
32bit int to uniquely id router
Comes from loopback or highest IP on router
What is “Default”?
Where to send traffic when there is no explicit route in the table
What happens in a BGP Announcement with NLRIs?
Each AS originates a set of NLRI.
These are exchanged
What if there are multiple paths to a given prefix?
BGP chooses the best and installs in the IP forwarding table
What are withdrawn routes composed of?
Network Prefix
Mask Length
iBGP mesh level
Full mesh, every router talks to every router
Route Reflectors
What do they do? What elements are required to prevent loops?
Pass on the best iBGP route updates
ORIGINATOR_ID and CLUSTER_LIST required to prevent loops.