cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
1190
Views
14
Helpful
5
Replies

EIGRP Question

Dev_update
Level 1
Level 1

Hi SME's ,

Have some questions , please help

Setup :  R1 -----128k-----> R2 -----64k----->R3 --- N

1) EIGRP Metric = (((10^7)/least-bandwidth) +cumulative-delay) * 256  , In R1 perspective , It would know the delay

and the BW assosicated with its own link , but  how does it know the bandwidth and delay assosicated with each

link towards the network N to calculate the least BW and the cumilative delay to calculate the eigrp metric ?

                         -------R2------

Setup :  N --- R1                   R3

                         -------R4-------

2) When the network N goes down , I know a triggered update will be sent out about the network N by the router R1 to R2 and R4 .Same

update gets relayed to R3 by R2 and my question is , Does the R3 will relay that on the other ways aswel , like R3 to R4 and back to R1  also from R3

to R2 and back to R1 . just dont understand ,please help

Thanks ,

Devis

5 Replies 5

Peter Paluch
Cisco Employee
Cisco Employee

Hi Devis,

1.) Regarding the bandwidth and delay components of the EIGRP metric, you have actually stated it yourself: the bandwidth is the minimal bandwidth along the prospective path while the delay is cumulative. When an EIGRP router advertises a network, say, R3, it indicates its own bandwidth and delay used to compute the metric in the Update packet. When R2 receives this update via the link to R3, it adds the delay of this link to the delay indicated in the Update packet, and subsequently, it compares the bandwidth indicated in the Update packet to the bandwidth of the link to R3, and it uses the smaller value.

The same happens when R2 forwards the Update to R1. This Update now contains the total cumulative delay from R2 to the destination network, and the minimal bandwidth between R2 and the destination network. R1 will increase the delay component by the delay of the link between R1 and R2, and it will take the smaller value from the tuple [bandwidth of the link between R1 and R2; bandwidth indicated in the Update].

This way, none of the routers needs to know individual delays and bandwidths of links towards a destination network because the delay is added at each hop and each router will use the appropriate lower bandwidth when computing the total metric.

2.) The actual sequence of steps in this topology will be different because of the diffusing update algorithm on which the EIGRP is based. When the network N goes down, R1 will send a query to R2 and R4 asking for a replacement route to the network. Because R2 and R4 used the R1 as their next hop (the successor) and now they see it has lost the route, both R2 and R4 know they have lost the route as well. Therefore, both R2 and R4 will propagate the query further to R3. Let's assume, for simplicity, that R3 receives the queries from R2 and R4 in almost the same time. The R3 will see that its next hops (successors) have lost the route to the destination network so it would like to propagate the query further but there is no other router to be asked for a backup route. So the R3 concludes that there is no replacement route and responds to R2 and R4 by saying that it does not know how to reach the network, and it removes the network from its routing table. The R2 and R4 receive this reply, and because they expect no other replies to arrive, they also see that there is no replacement route, so both R2 and R4 send their replies to R1, and they also remove the network from their routing tables. In the same way, the R1 will see that none of its neighbors is able to provide a replacement route, so it will simply delete the network from its routing table and the process is thereby finished.

Best regards,

Peter

Hi Peter ,

Great answer , need some clarifications  ,

Does the EIGRP updates contains the least bandwidth and cumulative-delay seperately by the neighbor router like AD for a given network ? .

So if a network N goes down , it is R1's responsibility to let all the other routers to know about it since the network N is directly connected to it .

What I understand is that in EIGRP this is been happening through query process , to let the other router to realize that the network is down . Is that correct ? .

Assume that the link comes up and could you please tell what happened as per my 2nd question ? .

I think I am bothering you too much .

Thanks ,

Devis

Hi Devis,

Thank you, you are welcome.

Does the EIGRP updates contains the least bandwidth and cumulative-delay
 seperately by the neighbor router like AD for a given network ?

If you are asking whether the EIGRP Update contains the four (actually five) metric components - bandwidth, delay, reliability, load, MTU - separate or whether they are already "mixed" together into a single metric value, then the answer is: the metric components are always advertised as separate values. It is up to the receiving routed to compute the resulting metric from the first four components but the EIGRP Update will always convey the individual components, never the resulting metric. That is actually one of the reasons why two EIGRP routers must absolutely agree on their K-values to compute the resulting metric in the same way, otherwise, each of them would arrive at a different result.

So if a network N goes down , it is R1's responsibility to let all the 
other routers to know about it since the network N is directly connected
 to it .

Basically, yes. In essence, a router that detects a metric change - both increase and decrease - must inform its neighbors about the event. It's as simple as that. A network going down is the same as having the metric of the network to grow to infinity, as both indicate that the network is unreachable.

What I understand is that in EIGRP this is been happening through query 
process , to let the other router to realize that the network is down . 
Is that correct ? .

There is a difference in EIGRP between an update process and a query process. A router sends an update when

  • the metric of a route has decreased
  • the metric of a route has increased but the router still has a valid successor to the destination

In essence, update means that the metric of a route has changed but the router sending the update still knows about a valid and usable route to the destination. Therefore, there is no need to ask the neighbors for an alternate path. It is sufficient to simply let them know about our own modified distance.

A query is sent when

  • the metric of a route has increased and there is neither successor nor feasible successor available anymore for the route
  • the metric of a route has increased and the next shortest path is provided by a neighbor that does not qualify as a feasible successor (i.e. it does not pass the feasibility condition check but still apparently provides the next shortest path)

In essence, when a router loses its next hop to a destination and is not sure which other neighbor shall it use to reach the destination without creating a routing loop, it has to ask by sending a query to its neighbors.

Assume that the link comes up and could you please tell what happened as per my 2nd question ?

The process is fairly simple in that case: the R1 will inform R2 and R4 about the newly connected network N using an Update packet. R2 and R4 will both learn about a new network, compute the metric (adding the delay and possibly modifying the bandwidth), store the route in their routing tables, and they will subsequently send the update further to R3, having modified the delay metric component and possibly the bandwidth metric component as well. R3 will receive these two (uncoordinated) updates, increase the delay, update the bandwidth if necessary, compute the metric, choose the best path and install it in its routing table. In addition, R2 and R4 will advertise this network back to R1 as unreachable (split horizon with poisoned reverse) and so will R3 towards its next hop (possibly to both R2 and R4 if all links are of the same speed and delay).

Best regards,

Peter

At least the answers to your questions would become rather long.

It is common practice to provide some links to related info on CCO which you can check for yourself.

For this subject, I would like to give you the link to the EIGRP page.

http://www.cisco.com/en/US/tech/tk365/tk207/tsd_technology_support_sub-protocol_home.html

On this page you will find a number of links with detailed technical information.

Likely you will find an answer to your questions here.

regards,

Leo

Dev_update
Level 1
Level 1

Thanks , you answers help a lot . Great ..

Devis