Border Gateway Protocol (BGP)
BGP Working Mechanism
v1.2
BGP Working Mechanism
BGP Messages
v1.2
BGP Messages
• Contains information required by BGP routers for maintaining
BGP session and performing routing updates.
• Each message is processed only after it is entirely received.
– Maximum size: 4,096 octets
• Must be supported by all BGP implementations
– Minimum size: 19 octets
• Message types:
– OPEN
– UPDATE
– NOTIFICATION
– KEEPALIVE
v1.2
BGP Message Types
Message Header:
• OPEN
– First message sent after TCP
connection establishment
– Contains:
• BGP version
• ASN of originating router Message Data:
• Hold Time
• BGP Router ID
• Capability codes (Optional)
– To be confirmed by KEEPALIVE
v1.2
BGP Message Types
Message Header:
• UPDATE
– Actual route updates
– Contains:
• Withdrawn Routes
Message Data:
• Path Attributes
• Network Layer Reachability Information (NLRI)
v1.2
BGP Message Types
Message Header:
• NOTIFICATION
– Sent when error occurred, contains
Error Code and Error Subcode
Error Symbolic Name Reference
Code
1 Message Header Error RFC4271 Section 6.1
Message Data:
2 OPEN Message Error RFC4271 Section 6.2
3 UPDATE Message Error RFC4271 Section 6.3
4 Hold Timer Expired RFC4271 Section 6.5
5 Finite State Machine Error RFC4271 Section 6.6
6 Cease RFC4271 Section 6.7
v1.2
BGP Message Types
Message Header:
• KEEPALIVE
– Does not contain data
– Sent to confirm OPEN, or to keep
HoldTimer from expiring
– As RFC4271’s suggested, Keepalive
Time is 30 seconds, and Hold Time is Message Data: (None)
90 seconds
• Keepalive Time and Hold Time are negotiable,
peers use the smaller values
• Each vendor may implement different default
configuration
v1.2
BGP Working Mechanism
BGP Finite State Machine (FSM)
v1.2
BGP Finite State Machine (FSM)
• BGP peer undergoes several state changes in its life cycle.
– IDLE
– CONNECT
– ACTIVE
– OPENSENT
– OPENCONFIRM
– ESTABLISHED
• During each state, peers must send and receive messages,
process message data, and initialize resources before
proceeding to the next state.
v1.2
BGP Finite State Machine (FSM)
• IDLE
ACTIVE CONNECT
– Initializes resources
– Resets ConnectRetryTimer OPEN OPEN
• RFC4271 suggested 120 seconds as the NOTIFICATION
OPENSENT IDLE
default value for ConnectRetryTime
– Initiates a TCP connection with its KEEPALIVE NOTIFICATION
configured BGP peer, and listens for a
TCP connection from its peer OPENCONFIRM
NOTIFICATION
• If no error, changes peer’s state to CONNECT
KEEPALIVE
• If error occurred, peer remains in IDLE state
ESTABLISHED
v1.2
BGP Finite State Machine (FSM)
• CONNECT
ACTIVE CONNECT
– Waits for successful TCP session
– Sends OPEN to peer OPEN OPEN
• If no error, changes peer’s state to OPENSENT NOTIFICATION
OPENSENT IDLE
• If error occurs, changes peer’s state to ACTIVE
• If ConnectRetryTimer expired, keeps peer in KEEPALIVE NOTIFICATION
CONNECT state and resets ConnectRetryTimer,
then tries a new TCP three-way handshake
OPENCONFIRM
• If something else happens, moves peer back to NOTIFICATION
IDLE state KEEPALIVE
ESTABLISHED
v1.2
BGP Finite State Machine (FSM)
• ACTIVE
ACTIVE CONNECT
– Unable to establish a successful TCP
session OPEN OPEN
– Tries to restart another TCP session with NOTIFICATION
OPENSENT IDLE
the peer
• If successful, sends an OPEN to the peer and KEEPALIVE NOTIFICATION
changes peer’s state to OPENSENT
• If unsuccessful, changes peer’s state to IDLE OPENCONFIRM
NOTIFICATION
• If ConnectRetryTimer expires, moves peer
back to CONNECT state KEEPALIVE
ESTABLISHED
v1.2
BGP Finite State Machine (FSM)
• OPENSENT
ACTIVE CONNECT
– OPEN has been sent to peer
– Waits for OPEN from peer OPEN OPEN
– Checks validity of the received OPEN NOTIFICATION
OPENSENT IDLE
• If there is no error, sends KEEPALIVE message
and changes peer’s state to OPENCONFIRM KEEPALIVE NOTIFICATION
• If error occurs due to mismatched OPEN
between peers, sends NOTIFICATION and OPENCONFIRM
change peer’s state to IDLE NOTIFICATION
• In case TCP session fails, moves peer back to KEEPALIVE
ACTIVE state
ESTABLISHED
v1.2
BGP Finite State Machine (FSM)
• OPENCONFIRM
ACTIVE CONNECT
– Waits for a KEEPALIVE or
NOTIFICATION from the peer OPEN OPEN
• Upon receipt of peer’s KEEPALIVE, changes
NOTIFICATION
peer’s state to ESTABLISHED OPENSENT IDLE
• If the HoldTimer expires or NOTIFICATION is
received, changes peer’s state to IDLE KEEPALIVE NOTIFICATION
OPENCONFIRM
NOTIFICATION
KEEPALIVE
ESTABLISHED
v1.2
BGP Finite State Machine (FSM)
• ESTABLISHED
ACTIVE CONNECT
– BGP peer adjacency is complete
– UPDATE is used for exchanging OPEN OPEN
reachability information NOTIFICATION
OPENSENT IDLE
• Initial full routing table exchange
• Incremental updates for later changes KEEPALIVE NOTIFICATION
– In case NOTIFICATION is received,
changes peer’s state back to IDLE OPENCONFIRM
NOTIFICATION
KEEPALIVE
ESTABLISHED
v1.2
BGP Working Mechanism
BGP Operations
v1.2
BGP Transport
• BGP operates by exchanging Network Layer Reachability
Information (NLRI).
– NLRI includes a set of BGP path attributes and one or more prefixes
which those attributes are associated
– NLRI is encapsulated inside the BGP UPDATE message
• Does not have own transport protocol.
• Utilizes TCP and runs on TCP port 179.
• BGP messages are exchanged over the TCP session.
v1.2
BGP Capabilities
• Capability codes indicate whether a BGP router is able to
accommodate particular capabilities.
– Advertised in OPEN message
• If received capability is not supported by remote peer, it sends
back a NOTIFICATION message.
• BGP routers attempt to peer without the unsupported capability.
• Commonly implemented capabilities:
– Route Refresh
– Multi-protocol Extension
– Support for 4-octet AS Number
v1.2
BGP Router ID
• Each router is identified by a unique 32-bit number written
in dotted decimal notation.
– Can be manually configured, or automatically selected by the router
• Cisco IOS: Highest Loopback IPv4 > Highest active interface IPv4
• Juniper Junos OS: Lowest Loopback IPv4 > Lowest physical interface IPv4
• MikroTik RouterOS: Lowest active interface IPv4
My Router ID My Router ID
is [Link]. is [Link].
[Link]/32, [Link]/128 [Link]/32, [Link]/128
Loopback0 Loopback0
.2 [Link]/24 .1 .2 [Link]/30 .1 .1 [Link]/24 .2
::2 [Link]/64 ::1 ::1 [Link]/127 ::0 ::1 [Link]/64 ::2
GigabitEthernet7 GigabitEthernet6 GigabitEthernet6 GigabitEthernet7
AS64512 AS65000
H1 [Link]/8 R1 BGP Session ISP1 [Link]/13 S1
[Link]/32 [Link]/32
v1.2
BGP Peer
• BGP does not perform auto-discovery for peers (neighbors).
• BGP peers are manually configured.
– Local peer address and ASN
– Remote peer address and ASN
– Authentication (if necessary)
I want to establish a I want to establish a
BGP session with BGP session with
[Link] from [Link] from
AS65000. AS64512.
[Link]/32, [Link]/128 [Link]/32, [Link]/128
Loopback0 Loopback0
.2 [Link]/24 .1 .2 [Link]/30 .1 .1 [Link]/24 .2
::2 [Link]/64 ::1 ::1 [Link]/127 ::0 ::1 [Link]/64 ::2
GigabitEthernet7 GigabitEthernet6 GigabitEthernet6 GigabitEthernet7
AS64512 AS65000
H1 [Link]/8 R1 BGP Session ISP1 [Link]/13 S1
[Link]/32 [Link]/32
v1.2
BGP Network
• Indicates a BGP prefix that should be originated by the router.
• By default, the prefix is advertised only if corresponding route
is present in the routing table.
My BGP networks: My BGP networks:
• [Link]/8 (aggregation) • [Link]/13 (aggregation)
• [Link]/24 (subnet) • [Link]/24 (subnet)
• [Link]/32 (aggregation) • [Link]/32 (aggregation)
• [Link]/64 (subnet) • [Link]/64 (subnet)
[Link]/32, [Link]/128 [Link]/32, [Link]/128
Loopback0 Loopback0
.2 [Link]/24 .1 .2 [Link]/30 .1 .1 [Link]/24 .2
::2 [Link]/64 ::1 ::1 [Link]/127 ::0 ::1 [Link]/64 ::2
GigabitEthernet7 GigabitEthernet6 GigabitEthernet6 GigabitEthernet7
AS64512 AS65000
H1 [Link]/8 R1 BGP Session ISP1 [Link]/13 S1
[Link]/32 [Link]/32
v1.2
BGP Network
• Prefixes are usually subnet or aggregate routes instead of
individual host routes.
• IPv4 prefixes longer than /24 and IPv6 prefixes longer than
/48 won’t generally be accepted on the Internet.
My BGP networks: My BGP networks:
• [Link]/8 (aggregation) • [Link]/13 (aggregation)
• [Link]/24 (subnet) • [Link]/24 (subnet)
• [Link]/32 (aggregation) • [Link]/32 (aggregation)
• [Link]/64 (subnet) • [Link]/64 (subnet)
[Link]/32, [Link]/128 [Link]/32, [Link]/128
Loopback0 Loopback0
.2 [Link]/24 .1 .2 [Link]/30 .1 .1 [Link]/24 .2
::2 [Link]/64 ::1 ::1 [Link]/127 ::0 ::1 [Link]/64 ::2
GigabitEthernet7 GigabitEthernet6 GigabitEthernet6 GigabitEthernet7
AS64512 AS65000
H1 [Link]/8 R1 BGP Session ISP1 [Link]/13 S1
[Link]/32 [Link]/32
v1.2
BGP Best Path
• Best path is the path that BGP selected to use in RIB.
• BGP uses path attributes to determine best path.
– Administrator influence the selection process by routing policy
– Best paths might not be the shortest path, but the most suitable path based on
the routing policy
• By default, BGP installs single best path for each destination.
• BGP propagates only the best path to the peers.
• BGP Multipath is a feature that allows BGP to install multiple best
paths when they have the same metrics.
– For load sharing over multiple next hops
v1.2
BGP Session Establishment
OPEN with 4-byte ASN Capability
NOTIFICATION: Unsupported Capabilities
AS65000 OPEN without Capability AS65001
KEEPALIVE
• Sends OPEN to peer after TCP three-way handshake.
• Peer replies NOTIFICATION if capabilities unsupported.
• Resends OPEN without unsupported capabilities.
• Peer replies KEEPALIVE if OPEN is acceptable.
• KEEPALIVE is sent periodically for maintaining the session.
v1.2
BGP Updates
Initial Full Update
AS65000 Subsequent Incremental Update AS65001
• Initial full update upon BGP session establishment.
• Subsequent incremental updates after initial full update.
– When new prefixes are being advertised
– When existing prefixes are being updated
– When existing prefixes are being withdrawn
v1.2
Route Refresh Capability
Route Refresh Request
AS65000 Full Update on Request AS65001
• Requests peer to resend full BGP update.
– RFC2918: Route Refresh Capability for BGP-4
• ROUTE-REFRESH message:
– AFI: Address Family Identifier
– Res.: Reserved field, set to 0 by the sender and ignored by the receiver
– SAFI: Subsequent Address Family Identifier
v1.2
