EIGRP automatically load-balances across links of equal cost. Whether the traffic is sent on a per-destination or round-robin basis depends on the internal switching within the router. It is possible to configure EIGRP to load-balance across unequal paths using the variance command. This command allows the administrator to identify by the use of the multiplier parameter the metric scope for including additional paths.

The command structure is shown here: variance multiplier

The multiplier is a number that ranges from 1 to 128. The default is 1, which allows for equal-cost load balancing. If the number is higher, it will multiply the best cost or metric value for a path by the number stated as the multiplier. All paths to the same destination that have metrics within this new range are now included in load balancing. The amount of traffic sent over each link is proportional to the metric for the path.

For example, the route to network A has four paths to it from router F and the best path gave a metric value of 10. The available routes shown in Figure 7-11 reflect these paths:

F to E to A = 30 F to D to B = 15 F to C to B = 15 F to C to G = 10

 E A -—, r" -—, r" Vc —-4 Vc —-u / 15 15 ,— / -—, r" -—, r" Vr —~À Vr i Vr Network A 5 -c —~t -c "—A C 5 G If the variance command was configured with a variance (multiplier) of 2; then the best metric is 10 x 2 = 20. These paths would all load-balance traffic from Router F to Network A: F to D to B = 15 F to C to B = 15 F to C to G = 10 One and a half packets would be sent across the path F to C to G for every one packet sent across the other two available paths.

NOTE The router cannot really send only one and a half packets and then switch the link, because the routers can't stop transmitting after they've started sending a packet. From a practical perspective, two packets are transmitted. This the same case for custom queuing.

NOTE Only those paths that are in the topology table as FS are eligible to be included in the variance command. Also, appreciate that the example and figure shown are highly simplified for the purpose of explanation.