CCNP - ospf costs/metrics

Author

ospf costs/metrics

2wired

2001-08-09, 8:34 pm

Just so I understand ospf costs/metrics correctly...

For an example....I have a remote WAN router with two PVC's. Both connections are 56k. One PVC connects to a host circuit @ 1024K and the other PVC connects to a different host circuit @ 512K. Because these connections are serial, without a bandwidth or cost statement on the host or remote interfaces will ospf see both circuits as 1536k links and load balance? With adding a bandwith statement on each host side, I assume ospf traffic will perfer the 1024 link? Then what if I add a higher cost (say above 1785) on the remote circuit that connects back to the 1024k host, will it then perfer the 512k?

Sorry about all of the questions, I am just somewhat confused.

MadChef

2001-08-10, 5:12 am

I believe you're generally right on all accounts. If no bandwidth statements are configured, the router will assume that bandwidth=port speed of the frame connection which is 1536k in your case. OSPF will treat these links as equal and will load balance across them so long as the cost to the destination network is equal. (It would be different if one PVC was "closer" to the destination network than the other and therefore wouldn't loadbalance because the ultimate cost of the route would be better on one pvc than the other.
The cost of a route is the sum of the cost on all outgoing interfaces to a route. So I believe that if you reconfigure the cost on the remote end, you will affect traffic coming from the remote end, but not going to the remote end.

Hope I've made some sense and haven't gotten too much wrong.

MadChef

tmadspiter

2001-08-10, 3:53 pm

No, they are seen as 2 different speeds with a lmi type of cisco(lmi). Annexd(ansi) is the other lmi type in which you have to set the bandwidth statement on the link. Just some helpful info, i am a frame relay eng., it one of the hardest wan connection to explain to some people, there is a lot of traffic shapping and such you can do with frame relay.

2wired

2001-08-10, 6:04 pm

That is an interesting point...I pull interface stats from mrtg. I could not figure out why almost all traffic was taking the 1024 link while only small amounts of ospf management traffic was seen on the 512 link, since there wasn't a bandwidth statement on the host or remote. Once a high cost was added (say to the 512 circuit)...then I actually saw a shift where the bulk of the data was still taking the 1024, however data started to show on the 512 circuit when the 1024 was hitting higher utilization.

Thanks everyone...from some other feedback I am ripping out the old bandwidth statements and using costs to shape the flow.

Sponsored Links