EIGRP Routing Question

[msheehan]

I am in the process of moving my Frame network from one provider to another. We have a 12mb Frame port with our new provider compared to T1's from our old provider. When I bought up our new providers circuit, some of our traffic started router through this circuit through other offices rather than the directly connected T1. I've added bandwidth statements to all of my interfaces to make the all equal. Now my traffic routes correctly. However, below is the output of a traceroute and "show ip route" to one of my offices. Is this correct? Is this load balancing now? It only seems to do this for the offices that are now connected to the new circuit.

yourrouter#show ip route myrouter

Routing entry for 192.10.239.48/30

Known via "eigrp 10", distance 90, metric 2809856, type internal

Redistributing via eigrp 10

Last update from 192.10.240.146 on Serial1.2, 01:18:43 ago

Routing Descriptor Blocks:

* 192.10.240.137, from 192.10.240.137, 01:18:43 ago, via Serial0.1

Route metric is 2809856, traffic share count is 1

Total delay is 45000 microseconds, minimum bandwidth is 1544 Kbit

Reliability 255/255, minimum MTU 1500 bytes

Loading 73/255, Hops 2

192.10.240.146, from 192.10.240.146, 01:18:43 ago, via Serial1.2

Route metric is 2809856, traffic share count is 1

Total delay is 45000 microseconds, minimum bandwidth is 1544 Kbit

Reliability 255/255, minimum MTU 1500 bytes

Loading 20/255, Hops 2

yourrouter#traceroute myrouter

Type escape sequence to abort.

Tracing the route to myrouter (192.10.239.49)

1 192.10.240.137 60 msec

192.10.240.146 68 msec

192.10.240.137 68 msec

2 192.10.240.113 136 msec

192.10.240.102 104 msec

192.10.240.113 136 msec

[mazhar71]

Hi,

My explaination may not fit you directly but load balancing depends on 2 factors mainly if you are not using policy routing or similar things.

1 - The routing table must see all available routes.

2 - Switching methods must be carefully choosen.

Switching method is important and may change. Basically 2 methods are available : per packet , per flow.

If you choose per packet, this is CPU sensitive process ( when per packet cef is not used ). But it really balance traffic equally with respect to the traffic share count of the routes.

If you choose per flow ( it is acctually per destination while not using cef ) traffic is shared by routing paths. If your traffic tends to flow to the same destination , the result may not be good as we consider load balancing.

I hope this is clear

Regards

[aacole]

I've had similar experiencies, you expect traffic will automatically flow over the fastest path, but it dosent always work out that way. Way I set up my FR networks using EIGRP is to use bandwidth=CIR on each PVC. Then if I want to influence routing I use the delay command on the PVC's that I want to make less attractive, increasing the value so that EIGRP picks my chosen route as a better path.

From your example the delay on int S1.2 is 45000 microseconds, what is the delay value on the new 12 meg link?

Remember also, if no bandwidth statement is set the link is considered to be T1 speed.