RIP and EIGRP

[RobBryson]

When installing a new pair of routers I initially was running RIP. I've now found the need to load balance over unequal cost paths and have turned on EIGRP.

Cisco-GP#show ip route

Codes: C - connected, S - static, R - RIP, M - mobile, B - BGP

D - EIGRP, EX - EIGRP external, O - OSPF, IA - OSPF inter area

N1 - OSPF NSSA external type 1, N2 - OSPF NSSA external type 2

E1 - OSPF external type 1, E2 - OSPF external type 2

i - IS-IS, su - IS-IS summary, L1 - IS-IS level-1, L2 - IS-IS level-2

ia - IS-IS inter area, * - candidate default, U - per-user static route

o - ODR, P - periodic downloaded static route

Gateway of last resort is 192.168.3.1 to network 0.0.0.0

C 192.168.4.0/24 is directly connected, FastEthernet0

C 192.168.2.0/24 is directly connected, Serial0

R 192.168.32.0/24 [120/1] via 192.168.3.1, 00:00:12, Ethernet0

[120/1] via 192.168.2.1, 00:00:04, Serial0

C 192.168.3.0/24 is directly connected, Ethernet0

S* 0.0.0.0/0 [1/0] via 192.168.3.1, Ethernet0

[1/0] via 192.168.2.1

Shouldn't the EIGRP routes be in the routing table due to their lower administrative distance? Obviously I'm missing something here and your help is appreciated.

[rwiesmann]

Hi

There are no EIGRP Route in this Routing Table.

Looks like you do not have a eigrp setup correctly.

Do you have an eigrp neighborship?

router#show ip eigrp neigbor

Regards

Roger

[RobBryson]

This is from one router...

Cisco-GP#show ip eigrp neighbor

IP-EIGRP neighbors for process 324

H Address Interface Hold Uptime SRTT RTO Q Seq

(sec) (ms) Cnt Num

1 192.168.3.1 Et0 13 17:40:59 1 4500 0 2

0 192.168.2.1 Se0 13 17:40:59 1 4500 0 1

This is from the second router...

Cisco-SHR#show ip eigrp neighbor

IP-EIGRP neighbors for process 324

H Address Interface Hold Uptime SRTT RTO Q Seq

(sec) (ms) Cnt Num

1 192.168.3.12 Et0 11 17:45:58 600 3600 0 2

0 192.168.2.2 Se0 14 17:45:59 1 3000 0 1

[rwiesmann]

Hi

Did you also give the network statement for the network 192.168.32.0 under the eigrp process of the router Cisco-SHR. Because it looks like this is the only route which you should get via eigrp but is still advertised via rip.

Do you have only two routers in this network?

Regards

Roger

[RobBryson]

Hello Roger,

Yes, there are only two routers in this network, and no, I didn't have a "network 192.168.32.0" statement under the eigrp process. The router Cisco-SHR lives in the 192.168.32.0 net and Cisco-GP lives in 192.168.4.0. Should Cisco-GP have only a "network 192.168.32.0" and Cisco-SHR only a "network 192.168.4.0" statement? I guess I'm very confused about which network each router needs to know about, the remote network or the local one.

[rwiesmann]

Hi

Its not so hard...just put the a network statement for the network of the line on which you have the eigrp neighbor under the eigrp process. This will establish the eigrp neighborship. This is what you have configured right now. But you also have other interfaces on your router with other subnet configured, like the 192.168.32.0. So if you like this network to be advertised by eigrp you also have to place a network statement under the eigrp process for it.

You now have the option to make this interface an passiv interface so that no eigrp neigborship can be built on that interface.

There are also other ways like redistribution to bring this network into eigrp.

Hope that helps

Roger

[RobBryson]

Thanks for your help so far Roger, eigrp now seems to be working. I do have a couple of more questions however.

Cisco-GP#show ip eigrp topology 192.168.32.0

IP-EIGRP (AS 324): Topology entry for 192.168.32.0/24

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

Routing Descriptor Blocks:

192.168.3.1 (Ethernet0), from 192.168.3.1, Send flag is 0x0

Composite metric is (284160/28160), Route is Internal

Vector metric:

Minimum bandwidth is 10000 Kbit

Total delay is 1100 microseconds

Reliability is 254/255

Load is 1/255

Minimum MTU is 1500

Hop count is 1

192.168.2.1 (Serial0), from 192.168.2.1, Send flag is 0x0

Composite metric is (2172416/28160), Route is Internal

Vector metric:

Minimum bandwidth is 1544 Kbit

Total delay is 20100 microseconds

Reliability is 254/255

Load is 2/255

Minimum MTU is 1500

Hop count is 1

Am I correct in thinking that there is only one feasible successor since there is only one hop to the destination network? The other question involves the routing table.

Cisco-GP#show ip route 192.168.32.0

Routing entry for 192.168.32.0/24

Known via "eigrp 324", distance 90, metric 284160, type internal

Redistributing via eigrp 324

Last update from 192.168.3.1 on Ethernet0, 00:00:18 ago

Routing Descriptor Blocks:

* 192.168.3.1, from 192.168.3.1, 00:00:18 ago, via Ethernet0

Route metric is 284160, traffic share count is 1

Total delay is 1100 microseconds, minimum bandwidth is 10000 Kbit

Reliability 254/255, minimum MTU 1500 bytes

Loading 1/255, Hops 1

If both Ethernet0 and Serial0 are viable paths why aren't they both listed as routes? If Ethernet0 fails do I really have the serial link to fail over to? The ultimate intent here is to load-balance over both paths, Ethernet0 getting the bulk of the traffic, but trying to use as much of the total bandwidth as possible.

[ziutek]

I don't think you will see much improvement in throughput if you load balance over two links with such different speed. Even if you have a T1, 1.5 mbs, you will basically send one packet over the serial link for every 7-8 over the ethernet link. The processing power needed to put these packets back in the proper order would more than likely offset any increase in bandwidth you might experience.

[RobBryson]

Agreed, but the boss wants to see this happen so I'll show him it can be done then point out the downside.

I'd still like an answer to the questions in my previous post.

[ruwhite]

This is going to depend on the type of load sharing and the type of switching you are doing. If you are doing per packet load sharing, then you won't see any improvement in real throughput. If you are running CEF, you would have to configure per packet load sharing--and I wouldn't recommend that for the reason cited above. If you are process switching packets, you can only per packet load share (see my description in my other reply on this thread).

If you are CEF switching, per flow load sharing is the defualt, which means that all the packets in a given flow will be switched across the same link, preventing out of order packets.

:-)

Russ.W

[Harold Ritter]

Eigrp allows unequal cost loadbalancing to take place. You need to use the "variance" knob under Eigrp to specify a mutiplier factor between the metrics for which Eigrp will allow loadbalancing. By default this multiplier is 1. In your case, it would need the be about 8 (2172416/284160=7.64).

This should give you the loadsharing that you expect.

Hope this helps,

[rwiesmann]

Hi

The variace command is one way to go. The other thing is you could balance it out with the delay command on the interface but then the two paths have to have the same metric. So you just configure a

delay on the ethernet so the to links have the same

coomposite metric.

Hope that helps

Roger

[Harold Ritter]

The approach you're proposing is rather complex and don't provide you with unequal cost path loadsharing.

My 2 cents.

[Kevin Dorrell]

Just as a curiosity, when you are doing unequal cost load balancing using the variance command, does the load get shared equally between the unequal cost routes, or does it get shared in proportion to the route costs?

[ruwhite]

It gets shared in proportion to the metrics on the routes. If you look at the routing table:

2651A#sho ip route 10.1.1.0

Routing entry for 10.1.1.0/24

Known via "static", distance 1, metric 0

Routing Descriptor Blocks:

* 208.0.7.1

Route metric is 0, traffic share count is 1

208.0.6.1

Route metric is 0, traffic share count is 1

Note the "traffic share count" information. What happens is this:

-- The routing protocol installs multiple routes to the same destination in the routing table.

-- The routing table examiens the routes, and finds the one with the highest metric.

-- The routing table then walks the paths to this destination, dividing the metric for each path into the highest metric it just found.

-- The resulting number is placed in the "traffic share count."

This means the traffic share count is set in proportion to the metrics fo the various routes in the table to the same destination. Now, when process switching, there is a counter, or rather a pointer, that points to the path the next packet switched should take. Every time a packet is process switched, this pointer is moved to the next path in the routing table. If the traffic share count is, say, 2, then two packets have to be switched along this path before the pointer is moved to the next path.

I hope this makes sense.

Now, if you're using CEF, it's a little different. CEF builds a series of tables, one of which is the load share table, which is populated based on the load share counters in the routing table.

There's a good description of how all of this works in the book _Advanced IP Network Design_.

:-)

Russ.W