EIGRP metric changes

Answered Question
Apr 28th, 2009
User Badges:
  • Purple, 4500 points or more

All,


If I were to change my weights, does that only affect the eigrp neighbors, or do all my upstream routers need to be changed as well?


If I have:


R1 (connected to R2 and R3)


R2 (connected to R5)


R3 (connected to R6)


I change my delay on R1, would I need to change the metric on ALL of the routers? Since it goes under the process, I would think it would affect all of the routers, which in my mind is a bad thing. Why would I ever want to change delay instead of letting the protocol figure it out?


Thanks,

John

Correct Answer by Edison Ortiz about 7 years 12 months ago

I'm not sure what the delay is between these links by default (let's say 1000 for ease of math). So, I would change the k-value for delay on R1, R2, and R3 to keep the adjacencies up. Does that mean that all of my delay has been changed by three for each hop?


R1 (1000*3=3000) --> R2 (1000*3=3000) --R3 (1000*3=3000)?


Yes, but the composite metric will be shown under the EIGRP route based on the formula I posted before. The delay modification will alter the EIGRP composite metric value but the idea you've presented is technically correct.


Also, what happens if I have two interfaces on R1, and I don't want the k-value to be modified for the other router. R1 is connected to R4 also, but I want the delay to be 2 instead of 3. Is that possible?


No, it affects the whole EIGRP process, it's not a per interface modification.


As I stated before, for per-interface modification, simply use the delay command under the interface without altering the metric weights.


__


Edison.


Correct Answer by Giuseppe Larosa about 7 years 12 months ago

Hello John,


for per interface manipulation you just need to do what Edison suggested:


verify current delay with


sh int type x/y


conf t

int type x/y

delay new-value


this doesn't change the K-values


Hope to help

Giuseppe


  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 5 (5 ratings)
Loading.
Edison Ortiz Tue, 04/28/2009 - 09:46
User Badges:
  • Super Bronze, 10000 points or more
  • Hall of Fame,

    Founding Member

If I were to change my weights, does that only affect the eigrp neighbors, or do all my upstream routers need to be changed as well?


If you were to change any of the 'k' values, the neighbor will complain about mismatched 'k' values, covered in the documentation:


http://www.cisco.com/en/US/docs/ios/iproute/configuration/guide/irp_cfg_eigrp_ps6350_TSD_Products_Configuration_Guide_Chapter.html#wp1054128


I change my delay on R1, would I need to change the metric on ALL of the routers? Since it goes under the process, I would think it would affect all of the routers, which in my mind is a bad thing. Why would I ever want to change delay instead of letting the protocol figure it out?


If you alter the corresponding K value for delay, you need to change it in all neighboring routers. If you enter the delay command under the interface, you are basically tuning routing preference instead of using the bandwidth command.




HTH,


__


Edison.

Giuseppe Larosa Tue, 04/28/2009 - 10:30
User Badges:
  • Super Silver, 17500 points or more
  • Hall of Fame,

    Founding Member

Hello John,

as Edison has explained the K-values have to match in order to build and to mantain EIGRP neighboships.

It is like using authentication all devices have to agree on the K vector.


if you change delay on an interface that is a different matter it is just a form of tuning tweaking.


changing delay is actually a more safe way to change EIGRP routing decisions then changing bandwidth: the reason is that EIGRP metric is cumulative in delay but inverse proportional to lowest BW on path.


For this reason several people prefer to modify delay instead of changing BW settings.


Note:

even more selective action is achieved by using offset-list where you can change the delay value (by adding the offset ) for selected routes


Hope to help

Giuseppe


John Blakley Tue, 04/28/2009 - 10:45
User Badges:
  • Purple, 4500 points or more

How does delay have an effect if I have to match all routers to the same k value? If I set all of my routers like:


router eigrp 1

metric weights 0 0 0 1 0 0


I set my delay to one, but how does that help with calculation? Is there a different way of changing delay? Does the "delay" command on an interface do anything for this?



Edison Ortiz Tue, 04/28/2009 - 10:51
User Badges:
  • Super Bronze, 10000 points or more
  • Hall of Fame,

    Founding Member

John,


The link I posted explains how you derive the composite metric. The formula is:


EIGRP Metric = 256*((K1*Bw) + (K2*Bw)/(256-Load) + (K3*Delay)*(K5/(Reliability + K4)))


If you are only using the delay to form the composite metric, any delay value under the interface will be taken into account for routing preference.


The delay command under the interface modifies the default delay value from such interface. To find default values under each interface, the show interface command will provide such information.


HTH,


__


Edison.

John Blakley Tue, 04/28/2009 - 10:57
User Badges:
  • Purple, 4500 points or more

Edison,


I read the document, and I'm sorry if I'm not making my question clear. I know that the k-values need to match through all of the links, and if I change my default delay from 0 to 3, then I need to change it to 3 on all of my links. My question is that if I have say three links:


R1(56k) --> R2 (768k) --> R3


I'm not sure what the delay is between these links by default (let's say 1000 for ease of math). So, I would change the k-value for delay on R1, R2, and R3 to keep the adjacencies up. Does that mean that all of my delay has been changed by three for each hop?


R1 (1000*3=3000) --> R2 (1000*3=3000) --R3 (1000*3=3000)?


Also, what happens if I have two interfaces on R1, and I don't want the k-value to be modified for the other router. R1 is connected to R4 also, but I want the delay to be 2 instead of 3. Is that possible?


Oh, and is my math right up there? =)


Thanks!

John

Correct Answer
Giuseppe Larosa Tue, 04/28/2009 - 11:03
User Badges:
  • Super Silver, 17500 points or more
  • Hall of Fame,

    Founding Member

Hello John,


for per interface manipulation you just need to do what Edison suggested:


verify current delay with


sh int type x/y


conf t

int type x/y

delay new-value


this doesn't change the K-values


Hope to help

Giuseppe


Correct Answer
Edison Ortiz Tue, 04/28/2009 - 11:04
User Badges:
  • Super Bronze, 10000 points or more
  • Hall of Fame,

    Founding Member

I'm not sure what the delay is between these links by default (let's say 1000 for ease of math). So, I would change the k-value for delay on R1, R2, and R3 to keep the adjacencies up. Does that mean that all of my delay has been changed by three for each hop?


R1 (1000*3=3000) --> R2 (1000*3=3000) --R3 (1000*3=3000)?


Yes, but the composite metric will be shown under the EIGRP route based on the formula I posted before. The delay modification will alter the EIGRP composite metric value but the idea you've presented is technically correct.


Also, what happens if I have two interfaces on R1, and I don't want the k-value to be modified for the other router. R1 is connected to R4 also, but I want the delay to be 2 instead of 3. Is that possible?


No, it affects the whole EIGRP process, it's not a per interface modification.


As I stated before, for per-interface modification, simply use the delay command under the interface without altering the metric weights.


__


Edison.


John Blakley Tue, 04/28/2009 - 11:11
User Badges:
  • Purple, 4500 points or more

Edison,


I have to be honest....I had to reread your first post and I asked every question twice that you had already answered. :-) My apologies.


Thanks,

John

Edison Ortiz Tue, 04/28/2009 - 11:13
User Badges:
  • Super Bronze, 10000 points or more
  • Hall of Fame,

    Founding Member

It's ok, I'll smack you later on :)


Giuseppe Larosa Tue, 04/28/2009 - 10:59
User Badges:
  • Super Silver, 17500 points or more
  • Hall of Fame,

    Founding Member

Hello John,


let me use Edison's job with a little change to help make it clear


EIGRP Metric = 256*((K1*Bw') + (K2*Bw)/(256-Load) + (K3*Delay')*(K5/(Reliability + K4)))


where

Delay' = Sum of all delays on path [tens of microseconds]


BW' = 10^7 / min BW [kbps]


so cumulative in delay and inverse proportional to minimum BW on path.


By the way, EIGRP routes carry all these values (including MTU) in separated fields in the route data structure.


the delay field stores the cumulative delay on path, the BW field stores the miminum BW on path


the local node has only to add its own delay contribution to verify if its BW setting is higher then that in the route to calculate the metric from its point of view.


Hope to help

Giuseppe


Actions

This Discussion