BGP Concepts

BGP (Border Gateway Protocol) is a standardized exterior gateway protocol (EGP) defined in RFC 4271. Its primary purpose is to exchange routing information between autonomous systems (ASes) on the internet. An autonomous system is a collection of networks under a single administrative domain, identified by a unique 16-bit or 32-bit AS number. BGP is classified as a path-vector protocol, meaning it maintains the path (sequence of AS numbers) to reach a destination network. This path information is used to detect and prevent routing loops.

Unlike interior gateway protocols (IGPs) like OSPF and EIGRP, which are designed for fast convergence within a single AS, BGP is designed for scalability and policy-based routing across the global internet. BGP can handle hundreds of thousands of routes, whereas IGPs typically handle a few thousand. BGP uses TCP port 179 for reliable transport, ensuring that routing updates are not lost. This reliance on TCP makes BGP slower to converge than IGPs, but reliability is prioritized over speed.

BGP routers (speakers) establish TCP connections (port 179) with peers in other ASes (eBGP) or within the same AS (iBGP). Once the TCP session is established, the peers exchange complete routing tables. After the initial exchange, only incremental updates are sent. BGP uses four message types:

When a BGP router receives multiple routes to the same destination, it must choose the best path. The BGP best path selection algorithm is a step-by-step process that evaluates routes based on attributes. The default order is: 1. Prefer the route with the highest WEIGHT (Cisco proprietary, local to router). 2. Prefer the route with the highest LOCAL_PREF (local preference, used within AS). 3. Prefer the route that was originated locally via network or aggregate command. 4. Prefer the route with the shortest AS_PATH. 5. Prefer the route with the lowest ORIGIN type (IGP < EGP < incomplete). 6. Prefer the route with the lowest MED (Multi-Exit Discriminator). 7. Prefer eBGP over iBGP. 8. Prefer the route with the lowest IGP metric to the NEXT_HOP. 9. If both are eBGP, prefer the oldest route (to minimize route flapping). 10. Prefer the route from the peer with the lowest router ID. 11. Prefer the route from the peer with the lowest peer IP address.

BGP uses attributes to describe routes. Attributes are classified as well-known mandatory, well-known discretionary, optional transitive, and optional non-transitive. Key attributes include:

A BGP session goes through several states before becoming established:

Basic eBGP configuration example:

router bgp 65001
 neighbor 10.0.0.2 remote-as 65002
 network 192.168.1.0 mask 255.255.255.0

Verification commands: - show ip bgp summary – displays BGP neighbor status, uptime, and route counts. - show ip bgp – shows the BGP table. - show ip route bgp – shows BGP routes in the routing table. - show ip bgp neighbors – detailed neighbor information. - debug ip bgp updates – monitors BGP updates (use with caution).

BGP interacts with IGPs through route redistribution. Typically, BGP routes are redistributed into an IGP (e.g., OSPF) only for iBGP sessions to avoid full mesh requirements. BGP also supports route reflectors and confederations to scale iBGP. MPLS VPNs use BGP to transport VPNv4 routes. BGP is also used in data center environments for EVPN (Ethernet VPN).

BGP is designed for the internet's scale. As of 2025, the global BGP table has over 900,000 routes. BGP routers must have sufficient memory to store these routes. Route aggregation (summarization) reduces table size. BGP also supports route filtering using prefix-lists, AS-path access-lists, and route-maps.

A company with a single AS (e.g., 65001) connects to two ISPs (AS 100 and AS 200) for redundancy. BGP is configured on the border routers. The enterprise advertises its public IP prefix (e.g., 203.0.113.0/24) to both ISPs. Inbound traffic is influenced by setting MED or AS_PATH prepending. For example, to prefer ISP1, the enterprise may prepend its AS three times on routes sent to ISP2: set as-path prepend 65001 65001 65001. Outbound traffic is controlled by setting LOCAL_PREF higher for the preferred ISP. BGP timers are tuned: hold time increased to 240 seconds to avoid flapping during brief outages. Memory considerations: the BGP table from both ISPs may contain 900k+ routes; the router must have sufficient RAM (at least 2 GB for full tables).

A service provider uses BGP to transport VPNv4 routes between Provider Edge (PE) routers. Each PE runs iBGP with route reflectors to avoid a full mesh. The provider uses BGP communities to tag routes for different customers (e.g., community 100:100 for gold service). Route maps apply QoS or bandwidth policies based on communities. Misconfiguration: forgetting to set the next-hop-self on iBGP peers causes reachability issues. Common fix: neighbor x.x.x.x next-hop-self.

Modern data centers use BGP for EVPN (Ethernet VPN) to provide VXLAN control plane. Spine switches run BGP with leaf switches. Each leaf advertises MAC/IP addresses via BGP EVPN routes. This replaces traditional flooding and learning. Scalability: BGP EVPN can handle thousands of tenants. Misconfiguration: incorrect route-target import/export leads to black holes. Troubleshooting involves checking show bgp l2vpn evpn.