EIGRP Wide Metrics

EIGRP uses the bandwidth, delay, reliability, load and K values to calculate a composite cost metric. The problem with this metric is that it doesn’t scale for high bandwidth interfaces. The composite cost metric is calculated like this:

EIGRP composite cost metric = 256*((K1*Scaled Bandwidth) + (K2*Scaled Bandwidth)/(256 – Load) + (K3*Scaled Delay)*(K5/(Reliability + K4)))

In the formula above, you can see that the bandwidth is scaled. It’s done with the following formula:

Scaled bandwidth = (107/minimum bandwidth (Bw) in kilobits per second)

The delay is also scaled, using this formula:

Scaled Delay = (Delay/10)

By default, only K1 and K3 are enabled, only bandwidth and delay are used. This means the composite cost metric can be simplified to the following formula:

EIGRP composite cost metric = 256*(Scaled Bandwidth + Scaled Delay)

The scaled bandwidth formula is unable to differentiate between anything faster than 10 GigabitEthernet. The lowest delay that we can configure is 10 microseconds, that’s the delay GigabitEthernet interfaces offer. Anything above GigabitEthernet will also have a delay of 10 microseconds.

Here are some examples to help you understand this:

• GigabitEthernet:
  • Scaled bandwidth: 10000000 / 1000000 = 10
  • Scaled delay: 10 / 10 = 1
  • Composite metric: 10 + 1 * 256 = 2816
• 10 GigabitEthernet:
  • Scaled bandwidth: 10000000 / 10000000 = 1
  • Scaled delay: 10 / 10 = 1
  • Composite metric: 1 + 1 * 256 = 512
• 11 GigabitEthernet:
  • Scaled bandwidth: 10000000 / 11000000 = 0.9 (rounded to 0)
  • Scaled delay: 10 / 10 = 1
  • Composite metric: 0 + 1 * 256 = 256
• 20 GigabitEthernet:
  • Scaled bandwidth: 10000000 / 20000000 = 0.5 (rounded to 0)
  • Scaled delay: 10 / 10 = 1
  • Composite metric: 0 + 1 * 256 = 256
• 40 GigabitEthernet:
  • Scaled bandwidth: 10000000 / 40000000 = 0.25 (rounded to 0)
  • Scaled delay: 10 / 10 = 1
  • Composite metric: 0 + 1 * 256 = 256

Above, you can see that the scaled bandwidth of anything higher than 10 GigabitEthernet has a value that is below one and is rounded down to zero. The delay has a value of one for GigabitEthernet or faster interfaces.

Any interface above 10 GigabitEthernet will always have a composite cost metric of 256. To EIGRP, there is no difference between anything above 10 GigabitEthernet or 40 GigabitEthernet interfaces and this might cause undesirable equal-cost load balancing in your network.

The metric and its formulas as explained above is called the EIGRP classic metrics. To solve this problem, EIGRP supports wide metrics which uses 64-bit values instead of the 32-bit values that EIGRP classics metric uses. EIGRP wide metrics supports interfaces up to 4.2 terabits and uses a different composite cost metric formula. It has the following components:

• Throughput: this is the bandwidth, it uses a new scaled bandwidth formula.
• Latency: this is the delay, in picoseconds. 1000000000 picoseconds = 1 millisecond. It also uses a new latency scaling formula.
• Reliability: the same as with EIGRP classic metrics.
• Load: the same as with EIGRP classic metrics.
• MTU: the same as with EIGRP classic metrics.
• Hop Count: the same as with EIGRP classic metrics.
• Extended Metrics: these are currently not used but reserved for future extensions. There are three extended metrics as of this moment:
  • Jitter
  • Energy
  • Quiescent Energy

To add these extended metrics to the composite metric, a new K value was introduced called K6.

Let’s take a look at the wide metrics composite metric formula: