cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
360
Views
0
Helpful
5
Replies

CBWFQ effect on EIGRP route selection

patrick.hurley
Level 3
Level 3

I have remotes with two paths to the host destination. Both paths are T1. One path is one hop and the other path is two hops. The one hop path has half of its bandwidth allocated for video when there is video traffice using CBWFQ. A CCIE told me that because of the metrics, data will flow over the single hop path until video shows up and then will overflow to the two hop path. What mechanism causes that to happen?

1 Accepted Solution

Accepted Solutions

Erick Bergquist
Level 6
Level 6

EIGRP has a combined metric which includes bandwidth, load, delay, link reliability, and a 5th factor which is calculated on reliability. By default bandwidth and delay are figured into the metric. You can add the other values if you configure EIGRP to use them with the metric-weights command.

The third option in the metric-weights command is set to 1 to include load.

ie: metric-weights 0111000 (default is 010100)

The load is dynamic and changes depending on load on the interface, whereas the bandwidth and delay values are usually constant. This is what the CCIE meant when the traffic would use both links when traffic overloaded the primary line but load has to be enabled in the metrics for EIGRP to take it into account.

View solution in original post

5 Replies 5

lindros
Level 1
Level 1

Are you using T1-PRI? What is config?

Pete -

clear channel T1

Erick Bergquist
Level 6
Level 6

EIGRP has a combined metric which includes bandwidth, load, delay, link reliability, and a 5th factor which is calculated on reliability. By default bandwidth and delay are figured into the metric. You can add the other values if you configure EIGRP to use them with the metric-weights command.

The third option in the metric-weights command is set to 1 to include load.

ie: metric-weights 0111000 (default is 010100)

The load is dynamic and changes depending on load on the interface, whereas the bandwidth and delay values are usually constant. This is what the CCIE meant when the traffic would use both links when traffic overloaded the primary line but load has to be enabled in the metrics for EIGRP to take it into account.

ruwhite
Level 7
Level 7

CBWFQ has nothing to do with the paths EIGRP chooses--it purly metric based, just the bandwidth and delay, which are constants. Another poster noted below you can change the k values to include the reliability and load; although these are updated on the interfaces in real time, they are not taken from the interfaces in real time from eigrp, so configuring them won't help you much in the way of doing any sort of "dynamically choose the lower utilized link," or anything like that.

If the two links are far apart in metric, and you can't bring them to match, then configure variance, so EIGRP will install both paths in the routing table, and CEF will unequal cost load share over them.

:-)

Russ

CBWFQ and EIGRP has nothing to do in common when it comes to path seleciton. Both should work independently.

One thing i notice is that when CBWFQ is configured on a Frame-relay main interface (using service-policy command), the policy might get removed, following a router reload, if bandwidth statements are configured on a subinterface. You should be careful tweaking bandwidth for EIGRP under these conditions, better tweak other parameters such as delay. Anyway the above situation doesnt apply to your case though.

Sankar Nair
UC Solutions Architect
Pacific Northwest | CDW
CCIE Collaboration #17135 Emeritus