About Basic BGP
Cisco NX-OS supports BGP version 4, which includes multiprotocol extensions that allow BGP to carry routing information for IP multicast routes and multiple Layer 3 protocol address families. BGP uses TCP as a reliable transport protocol to create TCP sessions with other BGP-enabled switches.
BGP uses a path-vector routing algorithm to exchange routing information between BGP-enabled networking switches or BGP speakers. Based on this information, each BGP speaker determines a path to reach a particular destination while detecting and avoiding paths with routing loops. The routing information includes the actual route prefix for a destination, the path of autonomous systems to the destination, and additional path attributes.
BGP selects a single path, by default, as the best path to a destination host or network. Each path carries well-known mandatory, well-known discretionary, and optional transitive attributes that are used in BGP best-path analysis. You can influence BGP path selection by altering some of these attributes by configuring BGP policies. See the Route Policies and Resetting BGP Sessions section for more information.
BGP also supports load balancing or equal-cost multipath (ECMP). See the Load Sharing and Multipath section for more information.
BGP Autonomous Systems
An autonomous system (AS) is a network controlled by a single administration entity. An autonomous system forms a routing domain with one or more interior gateway protocols (IGPs) and a consistent set of routing policies. BGP supports 16-bit and 32-bit autonomous system numbers. For more information, see the Autonomous Systems section.
Separate BGP autonomous systems dynamically exchange routing information through external BGP (eBGP) peering sessions. BGP speakers within the same autonomous system can exchange routing information through internal BGP (iBGP) peering sessions.
4-Byte AS Number Support
BGP supports 2-byte or 4-byte AS numbers. Cisco NX-OS displays 4-byte AS numbers in plain-text notation (that is, as 32-bit integers). You can configure 4-byte AS numbers as either plain-text notation (for example, 1 to 4294967295), or AS.dot notation (for example, 1.0). For more information, see the Autonomous Systems section.
Administrative Distance
An administrative distance is a rating of the trustworthiness of a routing information source. By default, BGP uses the administrative distances shown in the following table .
Distance |
Default Value |
Function |
---|---|---|
External |
20 |
Applied to routes learned from eBGP. |
Internal |
200 |
Applied to routes learned from iBGP. |
Local |
200 |
Applied to routes originated by the router. |
![]() Note |
The administrative distance does not influence the BGP path selection algorithm, but it does influence whether BGP-learned routes are installed in the IP routing table. |
For more information, see the Administrative Distance section.
BGP Peers
A BGP speaker does not discover another BGP speaker automatically. You must configure the relationships between BGP speakers. A BGP peer is a BGP speaker that has an active TCP connection to another BGP speaker.
BGP Sessions
BGP uses TCP port 179 to create a TCP session with a peer. When a TCP connection is established between peers, each BGP peer initially exchanges all of its routes—the complete BGP routing table—with the other peer. After this initial exchange, the BGP peers send only incremental updates when a topology change occurs in the network or when a routing policy change occurs. In the periods of inactivity between these updates, peers exchange special messages called keepalives. The hold time is the maximum time limit that can elapse between receiving consecutive BGP update or keepalive messages.
Cisco NX-OS supports the following peer configuration options:
-
Individual IPv4 or IPv6 address—BGP establishes a session with the BGP speaker that matches the remote address and AS number.
-
IPv4 or IPv6 prefix peers for a single AS number—BGP establishes sessions with BGP speakers that match the prefix and the AS number.
-
Dynamic AS number prefix peers—BGP establishes sessions with BGP speakers that match the prefix and an AS number from a list of configured AS numbers.
Dynamic AS Numbers for Prefix Peers and Interface Peers
[/bookmap/topic/concept/concept/concept/conbody {"../../../nexus_3500/6_x/nexus_3548_unicast_routing_configuration_guide/maps_and_wrappers/g_configuring_advanced_bgp.xml"})Cisco NX-OS accepts a range or list of AS numbers to establish BGP sessions. For example, if you configure BGP to use IPv4 prefix 192.0.2.0/8 and AS numbers 33, 66, and 99, BGP establishes a session with 192.0.2.1 with AS number 66 but rejects a session from 192.0.2.2 with AS number 50.
Beginning with Cisco NX-OS Release 9.3(6), support for dynamic AS numbers is extended to interface peers in addition to prefix peers. See Configuring BGP Interface Peering via IPv6 Link-Local for IPv4 and IPv6 Address Families.
Cisco NX-OS does not associate prefix peers with dynamic AS numbers as either interior BGP (iBGP) or external BGP (eBGP) sessions until after the session is established. See the chapter for more information on iBGP and eBGP.
![]() Note |
The dynamic AS number prefix peer configuration overrides the individual AS number configuration that is inherited from a BGP template. For more information, see the chapter. |
BGP Router Identifier
To establish BGP sessions between peers, BGP must have a router ID , which is sent to BGP peers in the OPEN message when a BGP session is established. The BGP router ID is a 32-bit value that is often represented by an IPv4 address. You can configure the router ID. By default, Cisco NX-OS sets the router ID to the IPv4 address of a loopback interface on the router. If no loopback interface is configured on the router, then the software chooses the highest IPv4 address configured to a physical interface on the router to represent the BGP router ID. The BGP router ID must be unique to the BGP peers in a network.
If BGP does not have a router ID, it cannot establish any peering sessions with BGP peers.
BGP Path Selection
Although BGP might receive advertisements for the same route from multiple sources, BGP selects only one path as the best path. BGP puts the selected path in the IP routing table and propagates the path to its peers.
The best-path algorithm runs each time that a path is added or withdrawn for a given network. The best-path algorithm also runs if you change the BGP configuration. BGP selects the best path from the set of valid paths available for a given network.
Cisco NX-OS implements the BGP best-path algorithm in the following steps:
Procedure
Step 1 |
Compares two paths to determine which is better (see the Step 1—Comparing Pairs of Paths section). |
Step 2 |
Iterates over all paths and determines in which order to compare the paths to select the overall best path (see the Step 2—Determining the Order of Comparisons section). |
Step 3 |
Determines whether the old and new best paths differ enough so that the new best path should be used (see the Step 3—Determining the Best-Path Change Suppression section). |
Example
![]() Note |
The order of comparison determined in Part 2 is important. Consider the case where you have three paths, A, B, and C. When Cisco NX-OS compares A and B, it chooses A. When Cisco NX-OS compares B and C, it chooses B. But when Cisco NX-OS compares A and C, it might not choose A because some BGP metrics apply only among paths from the same neighboring autonomous system and not among all paths. |
The path selection uses the BGP AS-path attribute. The AS-path attribute includes the list of autonomous system numbers (AS numbers) traversed in the advertised path. If you subdivide your BGP autonomous system into a collection or confederation of autonomous systems, the AS path contains confederation segments that list these locally defined autonomous systems.
Step 1—Comparing Pairs of Paths
This first step in the BGP best-path algorithm compares two paths to determine which path is better. The following sequence describes the basic steps that Cisco NX-OS uses to compare two paths to determine the better path:
-
Cisco NX-OS chooses a valid path for comparison. (For example, a path that has an unreachable next-hop is not valid.)
-
Cisco NX-OS chooses the path with the highest weight.
-
Cisco NX-OS chooses the path with the highest local preference.
-
If one of the paths is locally originated, Cisco NX-OS chooses that path.
-
Cisco NX-OS chooses the path with the shorter AS path.
Note
When calculating the length of the AS path, Cisco NX-OS ignores confederation segments, and counts AS sets as 1. See the AS Confederations section for more information.
-
Cisco NX-OS chooses the path with the lower origin. The Interior Gateway Protocol (IGP) is considered lower than EGP.
-
7. Cisco NX-OS chooses the path with the lower multi-exit discriminator (MED).
You can configure a number of options that affect whether or not this step is performed. In general, Cisco NX-OS compares the MED of both paths if the paths were received from peers in the same autonomous system; otherwise, Cisco NX-OS skips the MED comparison.
You can configure Cisco NX-OS to always perform the best-path algorithm MED comparison, regardless of the peer autonomous system in the paths. See the Tuning the Best-Path Algorithm section for more information. Otherwise, Cisco NX-OS will perform a MED comparison that depends on the AS-path attributes of the two paths being compared:
a. If a path has no AS path or the AS path starts with an AS_SET, then the path is internal, and Cisco NX-OS compares the MED to other internal paths.
b. If the AS path starts with an AS_SEQUENCE, then the peer autonomous system is the first AS number in the sequence, and Cisco NX-OS compares the MED to other paths that have the same peer autonomous system.
c. If the AS path contains only confederation segments or starts with confederation segments followed by an AS_SET, the path is internal and Cisco NX-OS compares the MED to other internal paths.
d. If the AS path starts with confederation segments followed by an AS_SEQUENCE, then the peer autonomous system is the first AS number in the AS_SEQUENCE, and Cisco NX-OS compares the MED to other paths that have the same peer autonomous system.
Note
If Cisco NX-OS receives no MED attribute with the path, then Cisco NX-OS considers the MED to be 0 unless you configure the best-path algorithm to set a missing MED to the highest possible value. See the Tuning the Best-Path Algorithm section for more information.
e. If the nondeterministic MED comparison feature is enabled, the best path algorithm uses the Cisco IOS style of MED comparison. See the Tuning the Best-Path Algorithm section for more information.
-
8. If one path is from an internal peer and the other path is from an external peer, then Cisco NX-OS chooses the path from the external peer.
-
If the paths have different IGP metrics to their next-hop addresses, then Cisco NX-OS chooses the path with the lower IGP metric.
-
Cisco NX-OS uses the path that was selected by the best-path algorithm the last time that it was run.
If all path parameters in Step 1 through Step 9 are the same, then you can configure the best-path algorithm to compare the router IDs. See the Tuning the Best-Path Algorithm section for more information. If the path includes an originator attribute, then Cisco NX-OS uses that attribute as the router ID to compare to; otherwise, Cisco NX-OS uses the router ID of the peer that sent the path. If the paths have different router IDs, Cisco NX-OS chooses the path with the lower router ID.
When using the attribute originator as the router ID, it is possible that two paths have the same router ID. It is also possible to have two BGP sessions with the same peer router, and therefore you can receive two paths with the same router ID.
-
Cisco NX-OS selects the path with the shorter cluster length. If a path was not received with a cluster list attribute, the cluster length is 0.
-
Cisco NX-OS chooses the path received from the peer with the lower IP address. Locally generated paths (for example, redistributed paths) have a peer IP address of 0.
Paths that are equal after step 9 can be used for multipath if you configure multipath. See the Load Sharing and Multipath section for more information.
Step 2—Determining the Order of Comparisons
The second step of the BGP best-path algorithm implementation is to determine the order in which Cisco NX-OS compares the paths:
-
Cisco NX-OS partitions the paths into groups. Within each group Cisco NX-OS compares the MED among all paths. Cisco NX-OS uses the same rules as in the Step 1—Comparing Pairs of Paths section to determine whether MED can be compared between any two paths. Typically, this comparison results in one group being chosen for each neighbor autonomous system. If you configure the bgp bestpath med always command, then Cisco NX-OS chooses just one group that contains all the paths.
-
Cisco NX-OS determines the best path in each group by iterating through all paths in the group and keeping track of the best one so far. Cisco NX-OS compares each path with the temporary best path found so far and if the new path is better, it becomes the new temporary best path and Cisco NX-OS compares it with the next path in the group.
-
Cisco NX-OS forms a set of paths that contain the best path selected from each group in Step 2. Cisco NX-OS selects the overall best path from this set of paths by going through them as in Step 2.
Step 3—Determining the Best-Path Change Suppression
The next part of the implementation is to determine whether Cisco NX-OS will use the new best path or suppress the new best path. The router can continue to use the existing best path if the new one is identical to the old path (if the router ID is the same). Cisco NX-OS continues to use the existing best path to avoid route changes in the network.
You can turn off the suppression feature by configuring the best-path algorithm to compare the router IDs. See the Tuning the Best-Path Algorithm section for more information. If you configure this feature, the new best path is always preferred to the existing one.
You cannot suppress the best-path change if any of the following conditions occur:
-
The existing best path is no longer valid.
-
Either the existing or new best paths were received from internal (or confederation) peers or were locally generated (for example, by redistribution).
-
The paths were received from the same peer (the paths have the same router ID).
-
The paths have different weights, local preferences, origins, or IGP metrics to their next-hop addresses.
-
The paths have different MEDs.
BGP and the Unicast RIB
BGP communicates with the unicast routing information base (unicast RIB) to store IPv4 routes in the unicast routing table. After selecting the best path, if BGP determines that the best path change needs to be reflected in the routing table, it sends a route update to the unicast RIB.
BGP receives route notifications regarding changes to its routes in the unicast RIB. It also receives route notifications about other protocol routes to support redistribution.
BGP also receives notifications from the unicast RIB regarding next-hop changes. BGP uses these notifications to keep track of the reachability and IGP metric to the next-hop addresses.
Whenever the next-hop reachability or IGP metrics in the unicast RIB change, BGP triggers a best-path recalculation for affected routes.
BGP Prefix Independent Convergence
The BGP prefix independent convergence (PIC) edge feature achieves faster convergence in the forwarding plane for BGP IP routes to a BGP backup path when there is a link failure.
The BGP PIC edge feature improves BGP convergence after a network failure. This convergence applies to edge failures in an IP network. This feature creates and stores a backup path in the routing information base (RIB) and forwarding information base (FIB) so that when the primary path fails, the backup path can immediately take over, enabling fast failover in the forwarding plane. BGP PIC edge supports only IPv4 address families.
When BGP PIC edge is configured, BGP calculates a second-best path (the backup path) along with the primary best path. BGP installs both best and backup paths for the prefixes with PIC support into the BGP RIB. BGP also downloads the backup path along with the remote next hop through APIs to the URIB, which then updates the FIB with the next hop marked as a backup. The backup path provides a fast reroute mechanism to counter a singular network failure.
This feature detects both local interface failures and remote interface or link failures and triggers the use of the backup path
BGP PIC edge supports both unipath and multipath.
BGP Virtualization
BGP supports virtual routing and forwarding (VRF) instances. By default, Cisco NX-OS places you in the default VRF unless you specifically configure another VRF. For more information, see Configuring Layer 3 Virtualization.