EIGRP Unequal cost LB, Eigrp Variance

Lessons learned:

EIGRP allows load distribution among unequal paths

-controlled by the variance command

If feasible distance * variance > feasible successor, load balancing occurs.

--ONLY feasible successors are candidate for load balancing

-Automatically calculated traffic share count causes links to be used in ratio proportional to their composite metrics

--Actual load balancing still controlled by switching path

Eigrp is the only IGP that can do this over different metrics.

As long as the backup paths viable we take our feasible distance – end to end metric for the active route – multiply it by the variance value. It the result of the calculation is larger than the metrics of the feasible successor then load balancing will occur.

The router will automatically calculate the traffic share between the primary path and the backup path. Based on the total end-to-end metric across the links.

Variance command:

Under the eigrp process:

Router eigrp 500

#variance 128

Does not matter what the variance value is as long as its large enough to take in account the feasible successors

Setting the variance to 10 20 or 30 will not affect the traffic share on the links, it will only affect what feasible successor can be candidate for equal cost load balancing.

So If we have a variance value of 128 – we multiply that by the FD. Any FS who’s end to end metric is lower the calculated value – they will be used for load balancing.

R3#sh ip eigrp topology 192.168.12.0/24

IP-EIGRP (AS 500): Topology entry for 192.168.12.0/24

  State is Passive, Query origin flag is 1, 1 Successor(s), FD is 284160

So if I have a FD of a route of 284160 then I multiply that by 128 – the total is 36372480. If this number is higher than the path through a router who we want to share the load with, then load balancing should occur.

The actual load balancing is going to depend on the underlying switching. Ex if we’re using CEF its going to source / destination / and the TCP flow information to figure out what interface to actually send the traffic out.

If we turned CEF off you could then do LB on a per packet basis.