EIRGP Question

Unanswered Question

Here is the output for the EIGRP route to the same device. One path is 100MB and the other is 1.5 MB.


Any thoughts as to why the 1.5 path has a smaller metric (by about half)?


Thank You



Routing entry for 10.2.37.0/24

Known via "eigrp 1", distance 90, metric 5148416, type internal

Redistributing via eigrp 1

Last update from 10.2.0.253 on GigabitEthernet0/1.1, 00:06:50 ago

Routing Descriptor Blocks:

* 10.2.0.5, from 10.2.0.5, 00:06:50 ago, via GigabitEthernet0/1.1

Route metric is 5148416, traffic share count is 1

Total delay is 200110 microseconds, minimum bandwidth is 100000 Kbit

Reliability 255/255, minimum MTU 1500 bytes

Loading 27/255, Hops 2




Holly#sh ip route 10.2.37.0

Routing entry for 10.2.37.0/24

Known via "eigrp 1", distance 90, metric 2989312, type internal

Redistributing via eigrp 1

Last update from 10.99.6.2 on Tunnel6, 00:37:36 ago

Routing Descriptor Blocks:

* 10.99.6.2, from 10.99.6.2, 00:37:36 ago, via Tunnel6

Route metric is 2989312, traffic share count is 1

Total delay is 50110 microseconds, minimum bandwidth is 1500 Kbit

Reliability 255/255, minimum MTU 1500 bytes

Loading 1/255, Hops 2




  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
Loading.
rais Fri, 05/08/2009 - 12:21
User Badges:
  • Silver, 250 points or more

There appears to be a lot more latency on your GE than on your tunnel interface. In addition, loading on tunnel is less than on GE.


Thanks.

Harold Ritter Fri, 05/08/2009 - 12:21
User Badges:
  • Cisco Employee,

Jeff,


In EIGRP, the metric is calculated from the minimum BW and the total delay along the path to destination. The first path does have a higher total delay than the second path, which would explains the higher metric. Someone might have set the delay manually on one of the interfaces along the first path.


Regards

I can not find any manual configuration of delay. My understanding of delay is it uses a table of interface types to figure out the delay. The first router is a 1GB link then a 100MB. The second router is connected by 2 tunnel interfaces at 1.5 MB.


Would different version IOS have a differnt delay table?

So it looks like it is an IOS version issue. I put the second router on the same version as the rest and now look at the route info. No changes other then the IOS.


Holly#sh ip route 10.2.37.1

Routing entry for 10.2.37.0/24

Known via "eigrp 1", distance 90, metric 14509312, type internal

Redistributing via eigrp 1

Last update from 10.99.6.2 on Tunnel6, 00:03:26 ago

Routing Descriptor Blocks:

* 10.99.6.2, from 10.99.6.2, 00:03:26 ago, via Tunnel6

Route metric is 14509312, traffic share count is 1

Total delay is 500110 microseconds, minimum bandwidth is 1500 Kbit

Reliability 255/255, minimum MTU 1500 bytes

Loading 1/255, Hops 2



lamav Sat, 05/09/2009 - 09:14
User Badges:
  • Blue, 1500 points or more

Hi:


You have to understand how the metric for EIGRP is derived. It's not enough to look at the bandwidth alone and compare the two. That is where you're going wrong.


The metric for an EIGRP route is the following:


Metric = [256 * (10^7/BW-min) + 256 * (Accumulated Delay)]


The delay is factored into the metric calculation, so you cant just use bandwidth and call it a day.


10^7 is what is referred to as a reference bandwidth.


Minimum bandwidth refers to the interface that has the lowest configured bandwidth of all the interfaces in the path between the SOURCE ROUTER -- be careful, not the source host that sent the traffic -- and the destination network. The bandwidth calculations begin with the first hop router's forward-facing (outgoing) interface. In other words, it is not the bandwidth from the perspective of the end-device that sourced the traffic that must first be considered, but the router's outgoing interface itself.


If a router forwards a packet that it received from a source host on its ethernet interface out of its serial interface, on its way to the next hop, then it is that bandwidth that gets consided as part of the calculation, not the bandwidth of the ethernet interface that has the source host connected to it.


As for delay, it is cumulative and taken not on an interface-per-interface basis, but from a link perspective. Delay is a quantification of the time a packet will take to traverse the link, end-to-end. So, you do not add the delay of directly connected interfaces to each other, but just take one reading of the link.


All this having been said, what you need to do is view the router's topology table to get a read on the minimum bandwidth of each path, from source to destination, as well as the cumulative delay.


Apply the formula, and you will see why one metric is higher than the other.


HTH


Victor

Actions

This Discussion