EIGRP Wide Metrics

This topic is to discuss the following lesson:

Hi Rene,
I have a question. What is the difference between composite metric and vector metric in EIGRP?
I see both the metric when I issue the command “sh ip eigrp x.x.x.x”

I appreciate if you can visualize me the difference between them.

Regards,
Mallik

Hello Mallik,

Although I am not Rene, I hope I can help you out. From what I understand vector metrics are the metrics that in total make up the composite metric. For example Bandwidth, Delay, Load, Reliability, and MTU are the Vector Metrics. Together they are used in a formula to find the composite metric. This composite metric is used by EIGRP to decide the best paths to insert into the topology and routing table.

Below is a quote and link to a Cisco article that you might find helpful.

"EIGRP Composite Cost Metrics
The Enhanced Interior Gateway Routing Protocol (EIGRP) uses bandwidth, delay, reliability, load, and K values (various constants that can be configured by a user to produce varying routing behaviors) to calculate the composite cost metric for local Routing Information Base (RIB) installation and route selections. The EIGRP composite cost metric is calculated using the following formula:

EIGRP composite cost metric = 256*((K1Scaled Bw) + (K2Scaled Bw)/(256 – Load) + (K3Scaled Delay)(K5/(Reliability + K4)))

EIGRP uses one or more vector metrics to calculate the composite cost metric. The table below lists EIGRP vector metrics and their descriptions"

I hope this helps you understand EIGRP better!

Thanks,
Scott

2 Likes

Hello Scott

That’s absolutely correct. Each of the K values (K1, K2, K3, K4, K5) are vectors and the composite metric is the result of the stated formula.

Thanks!

Laz

Many thanks Scott, well explained…appreciate your help.

2 Likes

Good day Network Lessons Team!
I was investigating the formula of calculating short composite metric with the relic components from IGRP: Load, along with the Reliability.
A)
256*((K1Scaled Bandwidth) + (K2Scaled Bandwidth)/(256 – Load) + (K3Scaled Delay)(K5/(Reliability + K4)))

I saw a lot of sources where we can find the same formula A , but there is another one B (the last parameter (K5/(Reliability + K4) is out of parentheses)
B)
256 * [K1Scaled Bandwidth + K2Scaled Bandwidth/(256 – Load) + K3*Scaled Delay] * K5/(Reliability + K4)

They give us different results. I tested it in GNS3 and observe that formula B is right. What have I missed?

Hello Evgeny

Looking at the RFC that describes EIGRP, the formula indicated is indeed formula “A” as stated in your post. Can you share with us where you found formula “B”? Also, it is true that you will indeed get different results with the different formulas, but this is the case only if you take K4 and K5 into account. Can you share your specific configuration with us that you performed in GNS3 to help us in responding?

I look forward to hearing from you.

Laz

Hello! Nice to hear you!

Formula B
CCIE R&S v5.0 vol.1 fifth edition
Narbik and Peter are writers, page 364

Formula A

Any EIGRP configuration in GNS3 at Cisco 7200, 3745 with “metric weights 0 1 0 1 1 1” gives me metrics of routes, and their values show me that formula B was used.

Hello again Evgeny

I found the reference in the Cert guide and started reading, and I realized that the two formulas are actually the same. This is because, for multiplication, the order of operations doesn’t matter. For example, if you have

256 * (a+b+c) * d

this is exactly the same as

256 * [(a+b+c) * d]

which is also exactly the same as

(256a + 256b + 256c) * d

You can replace

  • a with (K1Scaled Bandwidth)
  • b with (K2Scaled Bandwidth)/(256 – Load)
  • c with (K3Scaled Delay)
  • d with (K5/(Reliability + K4))

and you have the same thing. So it seems that both formulas will indeed bring the same result, because they are actually the same. I think the number of brackets has made it more difficult to see.

I hope this has been helpful!

Laz

Lazaros,
if my eyes don’t let me down
the formula A from the topic is:
256 * (a + b + c * d)

the formula B from the book is:
256 * (a + b + c) * d

Hello Evgeny

You’re right, the equation in the Cisco documentation you linked to is indeed formula A as shown below:


I turns out that this equation is wrong. This is a typo. (I didn’t look at this specific equation closely enough before, sorry…)

The equation from the RFC 7868 which is shown below is correct, and agrees with the equation in the Cert Guide:

5.6.1.1.  Classic Composite Formulation

   EIGRP calculates the composite metric with the following formula:

   metric = 256 * ({(K1*BW) + [(K2*BW)/(256-LOAD)] + (K3*DELAY)} * (K5/(REL+K4)))

So to finally :stuck_out_tongue_winking_eye: answer your question correctly, the correct equation is B, and the specific Cisco documentation has a typo.

I hope this has been helpful!

Laz

Thank you for clarifying!

1 Like

Why 256 is the scaling value in classic EIGRP for composite metric ?
Why 65535 is the scaling value in Named EIGRP for composite metric ?
Can we change the RIB Scale in EIGRP ?

Hello Narad

These are simply applied by definition. The protocol designers and engineers have found these values work well for the needs of the routing protocol. Because the classic metric no longer fulfils the needs of many high performance networks, the formula has been reengineered to resolve those issues with the wide metrics formula. These values are simply a result of studies, tests, ad trial and error.

Yes you can. The command to do this is metric rib-scale in the address family configuration mode. More info about this command can be found here:

I hope this has been helpful!

Laz

Hi,

I’m confused as to how dividing a 64-bit number by 128 reduces it to a 32-bit number.

To test, I converted the maximum 64-bits to decimal, divided by 128, then converted the decimal result back into binary, and it was 56-bits, which is the same as 64 bits - 8 bits (the binary value of 128).

Thanks.

Hello Samir

Yes you are correct that using a rib-scale of 128 may still not be enough to scale down the metric so that it fits into the 32-bit value of the routing table. Whenever the value exceeds what can be accommodated, the metric installed is a metric of infinity.

The rib-scale value is 128 by default, and works for most EIGRP networks. For some extensively large networks, you may find that metrics of infinity will be seen more often. If this happens, you are able to modify the rib scale value. This can be done using the metric rib-scale command. For more info on this command take a look at this Cisco command reference.

I hope this has been helpful!

Laz

Hi Laz,

Thanks, makes sense and matches what I found doing labs. That is, I used the delay to make a route less desirable, but when I disabled the preferred route to confirm failover, the less desirable route was not installed in the routing table as the high delay resulted in an infinite metic in the eigrp toplogy table.

Thanks,

Sam

Hi Laz,

I think there is a typo at the beginning in the first formulae. Bracket is in the wrong location:

EIGRP composite cost metric = 256*((K1Scaled Bandwidth) + (K2Scaled Bandwidth)/(256 – Load) + (K3Scaled Delay)(K5/(Reliability + K4)))

There should be a bracket in between (K5/(Reliability + K4) and the rest of the calculation before it is all multiplied by 256 .

EIGRP composite cost metric = 256*((K1Scaled Bandwidth) + (K2Scaled Bandwidth)/(256 – Load) + (K3*Scaled Delay))(K5/(Reliability + K4))

Thanks,

Sam

1 Like

Hello, everyone!

I am revisiting everything on the ENARSI blueprint to ensure that I have a good understanding before I take the exam and I want your opinion on something. When it comes to the EIGRP classic and wide metrics formula, should I be able to understand and use both to calculate the metric?

Classic isn’t that bad but Wide can end up with some large numbers if I have to calculate it manually. :smiley:

That’s all, thank you!

Kind regards,
David

Hello David

Regarding your question, understanding both EIGRP classic and wide metrics is beneficial. However, you don’t necessarily need to be able to manually calculate the EIGRP wide metric for the exam. It’s indeed a complex formula and can result in large numbers, as you’ve noted.

In my personal opinion, I find it extremely unlikely that you will have a question that asks you to calculate the metric even for classic EIGRP let alone wide metrics EIGRP, so I wouldn’t worry about it.

What’s more important is understanding the difference between the two and when to use each one. Classic EIGRP metrics are based on bandwidth and delay, while wide metrics consider additional parameters like jitter and reliability. This makes wide metrics more accurate and versatile, especially for modern networks that handle diverse types of traffic.

Remember that in real-world scenarios, these calculations are done by the routers themselves. Your role as a network engineer is to correctly configure and troubleshoot the EIGRP settings rather than manually calculate the metrics.

Good luck with your study and your exam!

I hope this has been helpful!

Laz

If the lowest bandwidth of a path is used for K1 calculation, how does a router with several hops to a destination know the whole picture of the path? It is not like an OSPF area where this information is shared?

Is this something that just carries over adjacencies starting from the lowest bandwidth advertiser of a path?