manual neighbor configuration

gojericho0gojericho0 Member Posts: 1,059 ■■■□□□□□□□
Could someone give me a scenario where you would want to manually configure a neighbor when they are normally discovered through multicast?

I see that it changes the hellos to unicast when neighbors are manually, is the only reason to cut down on multicast traffic?

Comments

  • marlon23marlon23 Member Posts: 164 ■■□□□□□□□□
    You can be forced to do it on CCIE lab :)
    LAB: 7609-S, 7606-S, 10008, 2x 7301, 7204, 7201 + bunch of ISRs & CAT switches
  • gojericho0gojericho0 Member Posts: 1,059 ■■■□□□□□□□
    This may be a little out of the scope of BSCI, but I was curious.

    What about in non-broadcast networks such as frame-relay and for some reason its a full mesh or point-to-multipoint network? Pseudobroadcasts are performed, but they are processed swithed and have to do X amount of duplicate packets for X amount of links in the multipoint connection. This doesn't seem to scalable if there are numerous links.

    I know the best choice is possible is to break the multipoint links into p2p links. If for whatever reason this is not viable solution would it be best to make the ospf network a nonbroadcast network and then configure the neighbors manually (Thinking this could be very tedious without a scripting mechanisim)?
    Router(config-if)# ip ospf network broadcast
    

    OR

    Use Protocol Independent Multicast (PIM) to allow the traffic to be fast-switched and take less resources on the router?
    Router(config-if)#  ip pim nbma-mode
    
  • marlon23marlon23 Member Posts: 164 ■■□□□□□□□□
    Hi again,

    I have a scenario for you, when it is good to use it :)

    Sometimes you can have neighbors on switched network (over VLAN f.e.), for example 1 with 10Mb ethernet, one with 100Mb FastEthernet and another one with 1Gb Ethernet. If your router has 100Mb interface he will assing the cost for OSPF based on this interface, however this is not correct cost to reach another neighbors (10Mb & 1Gb). Solution to this is to specify nonbroadcast OSPF network type there and use neighbor commands with cost option to fix traffic load balancing.

    Another similar scenario could be frame-relay point-to-multipoint subinterface at hub and different CIRs at spokes, then you would also prefer to set up a point-to-multipoint nonbroadcast network type and set cost values based on CIR per neighbor.

    So as you can see, if you care, you'll find a scenarios when specific ospf network type can solve some "issues".
    LAB: 7609-S, 7606-S, 10008, 2x 7301, 7204, 7201 + bunch of ISRs & CAT switches
  • networker050184networker050184 Mod Posts: 11,962 Mod
    I'd just put the cost on the interface which would be much easier, but I guess if you are looking for a place to use the neighbor command it would do the trick...
    An expert is a man who has made all the mistakes which can be made.
  • gojericho0gojericho0 Member Posts: 1,059 ■■■□□□□□□□
    Thanks guys, I appreciate it. Just trying to see the big picture with some of this stuff.
  • CiscoCertsCiscoCerts Member Posts: 112
    marlon23 wrote:
    he will assing the cost

    lol, sorry I couldn't help myself
Sign In or Register to comment.