Nokia BGP Fundamentals for Services key terms

A router that runs BGP and exchanges routing information with other routers. It can originate routes, receive routes, and apply policy before advertising them onward.

A group of IP networks under one administrative authority that presents a single, consistent routing policy to the Internet. Each AS is identified by a unique AS number.

A numeric identifier assigned to an Autonomous System. The 16-bit form is classic; 4-byte AS numbers extend the space and are carried between old and new speakers via AS4_PATH handling.

A remote BGP speaker with which a local speaker has formed a session and exchanges UPDATE messages. Peers are configured explicitly with an address and an autonomous system number.

A BGP session between two routers in the same autonomous system. Routes learned over iBGP are not re-advertised to other iBGP peers, so full reachability requires a full mesh or route reflectors.

A BGP session between routers in different autonomous systems. eBGP is the mechanism by which routes cross administrative boundaries, and it normally requires directly connected or reachable peers.

The TCP connection between two BGP speakers over which routing updates flow. It moves through Idle, Connect, Active, OpenSent, OpenConfirm, and Established states during setup.

The well-known port on which BGP listens and to which peers connect. A session cannot reach the Established state if this port is blocked by a filter or firewall in either direction.

The finite state machine that governs session setup and teardown. It moves from Idle through Connect, Active, OpenSent, OpenConfirm, and Established, and falls back on error or hold-timer expiry.

The first message exchanged after the TCP connection is up. It carries the peer's AS number, hold time, BGP identifier, and optional capabilities such as multiprotocol and 4-byte AS support.

The message that carries route advertisements and withdrawals. It includes path attributes, a list of withdrawn prefixes, and the NLRI being announced.

A short periodic message sent to confirm that the peer is still alive. Its interval is normally one third of the negotiated hold time.

A message sent when an error is detected, carrying an error code and subcode. Sending or receiving one causes the session to close and the FSM to return to Idle.

The negotiated interval after which a session is declared dead if no KEEPALIVE, UPDATE, or NOTIFICATION arrives. The two peers agree on the lower of their advertised values.

The interval at which a speaker sends KEEPALIVE messages to keep the session alive. It is typically one third of the hold time, so a 90-second hold time gives a 30-second keepalive.

A 32-bit value, usually the highest loopback IPv4 address, that uniquely identifies a speaker. Two peers with the same identifier will reject the session.

Network Layer Reachability Information: the encoded prefix and length carried in an UPDATE. It tells the peer exactly which destination the accompanying path attributes describe.

A field attached to an NLRI that describes how the route was learned and how it should be treated. Attributes are well-known or optional, and transitive or nontransitive.

The ordered list of autonomous systems a route has traversed. It is used for loop prevention and as the primary path-selection criterion after weight and local preference.

Repeating one's own AS number in an advertised AS_PATH to make the route look longer and therefore less attractive to inbound traffic. It influences the remote side's best-path choice.

The well-known attribute that names the address a router should use to reach the advertised prefix. For iBGP-learned routes, it is typically the eBGP peer's address unless changed.

The resolved forwarding address for a BGP route. If the next hop is unreachable in the IGP or RIB, the route is not eligible for best-path selection or installation.

A configuration that rewrites the NEXT_HOP of routes advertised to an iBGP peer to the local router's address. It avoids unreachable external next hops inside an AS.

A well-known discretionary attribute used inside an AS to prefer one exit point over another. Higher values win, and it is not advertised to external peers.

Multi-Exit Discriminator: an optional nontransitive attribute that hints to a neighboring AS which entry point to prefer. Lower values are better, and comparison is limited by default.

The ordered comparison a router runs over candidate routes to one prefix. It weighs weight, local preference, AS_PATH length, origin, MED, eBGP over iBGP, IGP metric, and tie-breakers.

A vendor-specific, locally significant value that is checked before local preference. Higher weight wins, and it is never advertised to any peer.

An iBGP speaker allowed to reflect routes from one client to another, removing the need for a full iBGP mesh. It uses cluster IDs and originator IDs to prevent loops.

A value carried in the CLUSTER_LIST attribute that identifies a route-reflector cluster. A reflector drops routes whose cluster list already contains its own ID.

An attribute added by a route reflector recording the router that originally injected the route. A speaker that sees its own router ID here discards the route as a loop.

A technique that splits one autonomous system into smaller internal sub-ASes for iBGP scaling. Sub-AS numbers are stripped at the confederation border, so external peers see one AS.

A configuration construct that applies the same outbound policy and timers to many neighbors at once. It reduces configuration size and update-processing overhead on large routers.

A set of match and action rules that filters or modifies routes on import or export. It implements prefix lists, AS-path filters, communities, and attribute changes.

A filter that permits or denies routes by prefix and prefix length. It is commonly used to enforce inbound and outbound route policy at a peering edge.

A 32-bit tag attached to routes and used to signal policy intent between operators. Well-known communities include NO_EXPORT and NO_ADVERTISE; others are locally defined.

An extension that lets BGP carry reachability for address families other than IPv4 unicast. Capabilities are negotiated in OPEN, and each family is advertised separately.

A distinct set of routes carried over one BGP session, such as IPv4 unicast, IPv6 unicast, or VPNv4. Peers must negotiate support for each family they intend to use.

An extended community that identifies which VRFs should import a VPN route. It is attached on export and matched against import policy on remote PE routers.

A virtual routing and forwarding instance that keeps a separate routing table on a PE router. BGP VPN routes are imported into the VRF whose route targets match.

An address family that carries Ethernet MAC and IP reachability in BGP, typically using route types 2 and 5. It supports data-center and service-provider overlay designs.

A stability mechanism that penalizes a prefix each time its route flaps and suppresses it once the penalty passes a threshold. The penalty decays exponentially over time.

Repeated establishment and teardown of a session or repeated withdrawal and re-advertisement of a route. Flapping consumes CPU and can trigger dampening or peer resets.

A capability that lets a restarting router keep forwarding while BGP reconverges. Peers retain routes during the restart and mark them stale until new updates arrive.

The process by which all speakers reach a consistent view after a topology or policy change. It depends on timers, update propagation, and best-path recomputation.

An eBGP session between peers that are not directly connected, requiring the packets to traverse intermediate routers. It usually needs an IGP route and a TTL adjustment.

Shaping how traffic reaches a network by influencing remote best-path decisions. Common tools are AS_PATH prepending, MED, and community-based policy at the neighbor.

Choosing which exit a network uses for traffic leaving its AS. It is done locally with LOCAL_PREF, weight, and route policy rather than by influencing remote routers.