EIGRP metrics
EIGRP is a Cisco proprietary routing protocol, developed to be an alternative to the IETF OSPF Link state routing protocol and an advance on restrictive and limited Distance Vector protocols like RIP and IGRP.
EIGRP is unusual like is predecessor in that it’s metric is not a cumulative value of one single path value such as hop count or bandwidth but a collection of interface values . This collection of values goes to create a metric called a Composite.
The Composite metric is composed of the following values:
- Bandwidth
- Delay (Measured in 10s
- Reliability
- Load
- MTU (Maximum Transmission Unit)
Even though EIGRP can use all five of these metrics only minimum bandwidth and delay are used to calculate the path metric.
It must be stressed that minimum bandwidth is the minimum bandwidth of the complete path, which means the bandwidth value does not indicate in any way number of router hops or how many low bandwidth interfaces there are along the path to the destination. Delay unlike bandwidth is a cumulative value which increments as the route is propagated along the path.
Cisco do not recommend Using Reliability and Load since these values are variable and will change dynamically causing the EIGRP protocol to recalculate the routes using cpu cycles allowing the device down. The MTU is NOT used in the calculation of the metric.
Bandwidth:
EIGRP uses 50% of the configured interface bandwidth. You might want to change the bandwidth value if a different value so that EIGRP updates do not consume or swamp the interface unnecessarily
The Metrics are also referred to as K values, and it is these values under the EIGRP routing process which need to be altered to either turn off and on or cause one metric value to be preferred more than another metric value.
Bandwidth is inverse minimum bandwidth of the path in bits per second scaled by a factor of 2.56 * 10 12.
Delay:
Delay is calculated a cumulative value of the complete path. The delay is measured in tens of microseconds. Setting the interface delay to 1 will display the delay as 10us will represent 0.01ms, or setting the delay to 10 will display the delay as 100us will represent 0.10ms
The formula to calculate the metric is:
metric = [k1 * bandwidth + (k2 * bandwidth)/(256 - load) + k3 * delay]
If k5 is not equal to zero, you will need to perform the additional operation:
metric = metric * [k5/(reliability + k4)]
The k values setting are carried across in every EIGRP hello packet, if the K values are mismatched it will cause the EIGRP neighbour relationship to fail
The EIGRP metric is represented with a 32-bit field max EIGRP metric is decimal 4,294,967,040
Metrics are configured as follows:
R2(config)#router eigrp 1
R2(config-router)#metric ?
holddown Enable EIGRP holddown
maximum-hops Advertise EIGRP routes greater than <hops> as unreachable
weights Modify EIGRP metric coefficients
R2(config-router)#metric weights ?
<0-4293967295> Type Of Service (Only TOS 0 supported)
TOS was intended by cisco to be used to mark EIGRP messages so they may be handled by QoS
R2(config-router)#metric weights 0 ?
<0-4293967295> K1
K1 = Bandwidth, this K value represents Bandwidth which is by default set to 1
R2(config-router)#metric weights 0 1 ?
<0-4293967295> K2
K2 = Load, this K value represents Load, which is by default set to 0
R2(config-router)#metric weights 0 1 0 ?
<0-4293967295> K3
K3 = Load, this K value represents Delay, which is by default set to 1
R2(config-router)#metric weights 0 1 0 1 ?
<0-4293967295> K4
K4 = Load, this K value represents Reliability, which is by default set to 0
R2(config-router)#metric weights 0 1 0 1 0 ?
<0-4293967295> K5
K5= Load, this K value represents MTU, which is by default set to 0
R2(config-router)#metric weights 0 1 0 1 0 0
The eigrp K-value is used to set how much importance is given to a component of the eigrp metric. So, by default, eigrp is set so that the "importance" of all K values apart from from bandwidth and delay is zero because the k-value for those components is zero.
The "importance" of bandwidth and delay is, by default, 1 because those components of the metric have a K-value of 1.
If you wanted eigrp to consider bandwidth twice as important as delay, for example, you would make bandwidth's K-value twice as large as the delay's K-value.
|
