Transport Network Guide

Routing Protocol Concepts

7 min read

Routing Protocol Concepts

Interior gateway protocols differ fundamentally in how routers discover and share topology information. This chapter covers the distance-vector and link-state paradigms, explains where hybrid protocols like EIGRP and path-vector protocols like BGP fit within that taxonomy, and establishes the role of loopback interfaces as stable router identities in production networks.

Content

EIGRP

Enhanced Interior Gateway Routing Protocol (EIGRP) is an interior gateway protocol originally developed by Cisco [3]. It is frequently described as a hybrid protocol because it borrows characteristics from both the distance-vector and link-state families [10]. EIGRP uses the Diffusing Update Algorithm (DUAL) to achieve rapid convergence and sends only partial, triggered updates rather than periodic full-table exchanges [3, 10].

Each EIGRP router maintains three data structures:

graph LR
    Hello["Hello Packets"] --> NT["Neighbor Table<br/>(adjacencies)"]
    NT --> TT["Topology Table<br/>(all routes from all neighbors)"]
    TT -->|"DUAL selects best"| RT["Routing Table<br/>(best routes only)"]
    TT -->|"DUAL pre-computes"| FS["Feasible Successor<br/>(backup path)"]

    style NT fill:#ffd,stroke:#333
    style TT fill:#dff,stroke:#333
    style RT fill:#dfd,stroke:#333
    style FS fill:#fdd,stroke:#333

DUAL pre-computes a backup path called the feasible successor [3, 10]. When the primary (successor) route fails, the router promotes the feasible successor immediately without recalculating. This eliminates the count-to-infinity problem that plagues classical distance-vector protocols [10].

PropertyValueNotes
Metric componentsBandwidth, delay, reliability, loadOnly bandwidth + delay used by default [3]
AD (internal)90IGP-learned routes
AD (external)170Redistributed routes
AD (iBGP)200Deliberately higher than EIGRP — IGP always preferred

Deployment Scope

EIGRP was Cisco-proprietary for decades. Cisco published it as informational RFC 7868 in 2016, but real-world adoption outside Cisco-only environments remains negligible. Multi-vendor networks standardize on OSPF or IS-IS.

Deployment scope. EIGRP was Cisco-proprietary for decades. Cisco published it as informational RFC 7868 in 2016 [3], but real-world adoption outside Cisco-only environments remains negligible. Multi-vendor networks standardize on OSPF or IS-IS instead.

The core distinction between these two paradigms is the scope of each router’s knowledge about the network.

Distance Vector

In a distance-vector protocol, each router knows only the distance (a cost metric) and the vector (the next-hop neighbor) toward every destination. Routers periodically transmit their entire routing table to directly connected neighbors and accept what neighbors advertise without independent verification — a model sometimes called “routing by rumor.”

RIP is the canonical example: it uses hop count as its metric (maximum 15), sends full table updates every 30 seconds (RIPv1 broadcasts; RIPv2 multicasts to 224.0.0.9), and converges slowly after topology changes [4].

The fundamental weakness is the count-to-infinity problem [7]. When a link fails, routers exchange stale information in a loop, each incrementing the metric until it reaches the protocol maximum. Several mitigations exist — split horizon, route poisoning (poison reverse), and hold-down timers — but these are workarounds for a structural limitation, not solutions to it [7].

In a link-state protocol, each router independently discovers its own directly connected links and floods that information as Link-State Advertisements (LSAs) to every other router in the area [1, 8]. Every router therefore builds an identical copy of the full topology database and independently runs Dijkstra’s Shortest Path First (SPF) algorithm to compute the optimal path tree to every destination [1, 8]. This model is sometimes called “routing by map.”

OSPF is the most widely deployed link-state protocol [1]. Convergence is fast because only small, incremental LSAs are flooded when a change occurs — not the entire routing table [1, 8]. The trade-off is higher CPU and memory consumption, since every router stores and processes the complete topology.

Comparison

PropertyDistance VectorLink State
Topology knowledgePartial — neighbors onlyComplete — full map
Update mechanismFull table, periodicIncremental LSAs, event-triggered
AlgorithmBellman-FordDijkstra SPF [1]
Convergence speedSlow (minutes)Fast (seconds)
Loop preventionSplit horizon, poison reverse, hold-down timers [7]Inherently loop-free (consistent global view)
Resource demandsLow CPU and memoryHigher CPU and memory
ScalabilitySmall networksLarge enterprise and service-provider networks
Representative protocolsRIP, IGRP [4]OSPF, IS-IS [1, 5]

Positioning EIGRP and BGP

graph TD
    subgraph "Routing Protocol Classification"
        DV["Distance Vector<br/>(RIP, IGRP)"]
        HYB["Hybrid / Advanced DV<br/>(EIGRP)"]
        PV["Path Vector<br/>(BGP)"]
        LS["Link State<br/>(OSPF, IS-IS)"]
    end

    DV -->|"+ topology table<br/>+ triggered updates<br/>+ DUAL backup paths"| HYB
    DV -->|"+ full AS path<br/>+ policy attributes"| PV
    DV -.->|"fundamentally different<br/>knowledge model"| LS

    style DV fill:#f9d,stroke:#333
    style HYB fill:#ffd,stroke:#333
    style PV fill:#dff,stroke:#333
    style LS fill:#dfd,stroke:#333

EIGRP is structurally a distance-vector protocol — routers share reachability information with neighbors rather than raw topology [3]. However, it borrows link-state characteristics such as triggered updates, a topology table, and pre-computed backup paths [3, 10]. DUAL eliminates count-to-infinity entirely, which is why EIGRP earns the “hybrid” label.

BGP operates as a path-vector protocol, an evolution of the distance-vector concept. Instead of advertising a simple distance-plus-direction tuple, BGP advertises the complete AS path to every destination. Loops are inherently prevented because a router rejects any route whose AS path already contains its own AS number.

Intuition

Distance vector is like asking a stranger for directions: “the airport is three blocks that way.” You trust the answer but cannot verify it. Link state is like everyone sharing their section of a city map — every participant assembles the same complete map and independently finds the best route.

Loopback Interfaces

A loopback is a virtual interface that exists entirely in software. It has no physical port — just an IP address assigned to the router itself. The familiar 127.0.0.1 is the operating-system loopback (localhost); on routers, administrators create additional loopback interfaces (Loopback0, Loopback1, and so on) and assign them routable IP addresses, conventionally with a /32 mask [7].

The defining property of a loopback is permanence: it never transitions to a down state unless the entire router fails [7, 8]. Physical interfaces can go down due to cable faults, port failures, or administrative shutdowns. A loopback remains up as long as the router is running, making it the most stable identity a router can present to the network.

Use Cases

Use CaseWhy Loopback?What Breaks Without It
OSPF/BGP Router IDStable identity that never flaps [1, 8]Protocol picks highest physical IP — fragile if that interface fails
iBGP PeeringSession survives individual link failures via IGP reroute [7]Physical-IP peering drops when that specific link goes down
Management (SSH)Reachable via any surviving path to the routerSSH to physical IP unreachable if that interface is down
MPLS/VPN EndpointLSR router-id and tunnel endpoint for the VPN fabricMPLS label bindings and VPN tunnels break on interface failure
neighbor 1.1.1.1 remote-as 100
neighbor 1.1.1.1 update-source Loopback0

Key Insight

The /32 mask is standard because a loopback represents a single host address — there is no subnet behind it, only the router itself. A loopback gives every router a permanent, hardware-independent identity and is considered mandatory in production network designs.

See Also

  • BGP Fundamentals (01-bgp-fundamentals) — path-vector classification, AS_PATH, eBGP peering, administrative distance context
  • BGP Operations: Aggregation, iBGP, and Route Reflectors (03-bgp-operations) — iBGP full-mesh, route reflectors, loopback-based peering in practice
  • Tier-1 SP Architecture and L3VPN (SP-Architecture/01-tier1-sp-architecture-l3vpn) — OSPF/IS-IS as IGP underlay for MPLS/VPN fabric

References

  1. RFC 2328 — OSPF Version 2. J. Moy, IETF, April 1998. https://www.rfc-editor.org/rfc/rfc2328
  2. RFC 5340 — OSPF for IPv6. R. Coltun et al., IETF, July 2008. https://www.rfc-editor.org/rfc/rfc5340
  3. RFC 7868 — Cisco’s Enhanced Interior Gateway Routing Protocol (EIGRP). D. Savage et al., IETF Informational, May 2016. https://www.rfc-editor.org/rfc/rfc7868
  4. RFC 2453 — RIP Version 2. G. Malkin, IETF, November 1998. https://www.rfc-editor.org/rfc/rfc2453
  5. RFC 1195 — Use of OSI IS-IS for Routing in TCP/IP and Dual Environments. R. Callon, IETF, December 1990. https://www.rfc-editor.org/rfc/rfc1195
  6. ISO/IEC 10589:2002 — Intermediate System to Intermediate System intra-domain routeing information exchange protocol (IS-IS). https://www.iso.org/standard/30932.html
  7. J. Doyle & J. Carroll, Routing TCP/IP, Volume I, 2nd ed., Cisco Press, 2005.
  8. J. Moy, OSPF: Anatomy of an Internet Routing Protocol, Addison-Wesley, 1998.
  9. A. Martey, IS-IS Network Design Solutions, Cisco Press, 2002.
  10. R. Pepelnjak, J. Apcar, EIGRP for IP, Addison-Wesley, 2000.

Graph View