skip to content

What does an EIGRP router's composite metric measure along a path by default, and which carried values does it leave out?

level: juniorimportance: must knowfreq 34%

answer

  1. a weakest link and a running total
  2. inverse of the slowest link
  3. static per-interface delay, summed
  4. K1 and K3 on, the rest at zero

basics

~20 s

By default the EIGRP composite metric adds two terms: the inverse of the slowest link's bandwidth on the path and the sum of the outgoing interfaces' delays. Load, reliability, MTU and hop count travel with the route but are not weighed.

solid answer

~40 s

EIGRP advertises a vector for each route — minimum bandwidth, total delay, maximum load, minimum reliability, minimum MTU and hop count — and each router folds it into one number using weights called K-values. With RFC 7868's defaults, `K1 = K3 = 1` and `K2 = K4 = K5 = 0`, only two terms survive: `10^7` divided by the slowest link's bandwidth in kb/s, and the sum of the outgoing interface delays in tens of microseconds, the total multiplied by 256. Lower is better. The delay is a configured per-interface value, not a measured latency. MTU is carried but RFC 7868 says it is never part of the metric, and hop count is carried without being a metric term.

go deeper

for a junior

Recall the two default inputs: the slowest link's bandwidth and the total delay along the path. Lower metric wins.

for a middle

Explain how each vector component is combined hop by hop, minimum versus sum, and why the default K-values leave only bandwidth and delay in play.

for a senior

Point out that delay is a configured value, so the metric never reacts to congestion, and that this stability is exactly why the defaults leave load out.

for a principal

Frame the composite metric as a design choice: a stable, static ranking that every router computes identically, traded against any ability to steer around live congestion.

## Why EIGRP needed more than a hop count A **distance-vector** routing protocol learns routes from its neighbours and ranks them by a number called the **metric**; lower is better. The simplest metric is a hop count, and it has an obvious flaw: two hops across slow links beat three hops across fast ones. EIGRP — a distance-vector protocol that was one vendor's for years and was published as **Informational RFC 7868** in 2016 — replaces the hop count with a **composite metric**: a single number built from several properties of the path. To make that possible, every route EIGRP advertises carries a small **vector** of path properties rather than one finished number. Each router updates the vector as the route passes through it, and each router turns the vector into a composite metric using weights called **K-values**. ## The vector each route carries RFC 7868 §5.6.2.1 lists how each component is combined from the destination back towards the router: | Component | Combined hop by hop as | In the default metric? | |---|---|---| | Bandwidth | the **minimum** along the path | yes, weighted by `K1` | | Delay | the **sum** along the path | yes, weighted by `K3` | | Load | the **maximum** along the path | no, `K2 = 0` | | Reliability | the **minimum** along the path | no, `K4 = K5 = 0` | | MTU | the **minimum** along the path | never: RFC 7868 says MTU is not an attribute for calculating the metric | | Hop count | the **sum** along the path | no: carried, not a metric term | ## The default formula RFC 7868 defines the default weights as `K1 = K3 = 1` and `K2 = K4 = K5 = 0` (and `K6 = 0`, a weight that only the newer wide metrics use). With those values the classic formula collapses to two terms: ``` metric = 256 × ( 10^7 / minimum bandwidth in kb/s + sum of delays in tens of microseconds ) ``` - **The bandwidth term** is `10^7` divided by the slowest link's bandwidth in kilobits per second. It is an inverse: the slower the bottleneck, the bigger the term, and the worse the route. - **The delay term** is the sum of the delays of the **outgoing interfaces** along the path, counted in units of ten microseconds. - **The factor 256** is historical: EIGRP kept its predecessor protocol's composite formula and multiplied it by 256 to widen a 24-bit metric to 32 bits. - **Reliability drops out cleanly.** The full formula multiplies by `K5 / (REL + K4)`, but RFC 7868 defines that reliability quotient to be 1 whenever `K5` is 0, so the default metric is not zeroed by it. ## What the two terms mean in practice 1. **The bandwidth term is a weakest link.** It looks only at the slowest link on the path. Adding more fast hops does not change it, and one slow hop dominates however fast the rest are. 2. **The delay term is what makes path length count.** Every hop adds its interface delay, so among paths with the same bottleneck the one with less accumulated delay — usually the shorter one — wins. 3. **Delay is configured, not measured.** RFC 7868 describes it as an administrative value assigned per interface to represent the time across an unloaded path. A congested link does not report a higher delay. The per-interface default values come from the implementation, not from the protocol. 4. **The result is compared, then advertised onward.** The router keeps the route with the lowest composite metric and advertises the updated vector — not the composite number — to its own neighbours, who apply the same K-values. That last point is why the K-values are an adjacency matter: every router must turn the same vector into the same ranking. Which checks a new neighbour must pass belongs to EIGRP's neighbour formation; the reason behind this one is the metric's. ## Common mistakes - Treating the bandwidth term as a sum or an average of the links. It is the minimum. - Believing EIGRP measures latency. The delay is a static per-interface value. - Listing MTU as a metric input. It travels in the vector so the path's smallest MTU is known, and the RFC excludes it from the calculation. - Counting hops. The hop count is carried in the vector but is not one of the weighted terms. - Saying load and reliability are part of the default metric. They are carried, and their weights are zero.

  • If load and reliability are not used by default, why does EIGRP carry them in every route at all?
    Because the K-values can switch them on. RFC 7868 lets K2 weigh load and K4 and K5 weigh reliability, so the vector must carry them for any router to apply those weights. With the defaults they are carried but multiplied by zero, or, for reliability, replaced by a quotient of 1.
  • Two EIGRP paths share the same 100 Mb/s bottleneck; what decides between them?
    Only the delay term. The bandwidth terms are identical because both use the same minimum, so the path whose outgoing interfaces sum to less delay has the lower metric. With per-interface delays, that is usually the path with fewer or faster hops.

A convoy can travel no faster than its slowest truck, and every stop along the way adds waiting time. EIGRP's default metric is exactly that: the slowest link on the path plus the delay added at every hop.

saying these in an interview costs you the question

  • EIGRP adds up the bandwidth of every link on the path.
  • EIGRP measures each link's latency with probes and feeds that in.
  • MTU is one of the default composite-metric terms.
  • Load and reliability are always part of the metric.
  • A higher EIGRP metric means a better route.