Pages

4.05.2013

OSPF Network Types Part 7 - Multipoint Non Broadcast

Last post of the OSPF Network Type series!! I hope it was as good for you as it was for me.

The last option for your OSPF network type is Multipoint Non Broadcast. As mentioned in the Multipoint post Multipoint Non Broadcast has the ability to apply accurate cost to routes when calculating the metric on Multipoint link who have different bandwidth coming in from the spoke.



As usual here are the specifics before we dive in.

  • Uses unicast for hellos
  • No neighbor discovery
  • No DR/BDR election
  • Hello/Dead timers 30/120
  • Cisco Proprietary
We are going to modify the topology some.

Configuration will be the same as last time. No frame maps or neighbor statements to start. Only difference is we will add an additional link between R4 and R2 and add it to area 0 of OSPF.

First lets verify that R1 has learned all routes and can see the loopback of R4 (4.4.4.4)



Notice R1 has two routes installed for 4.4.4.4.

Now lets say that R2 was not setup with a standard T1 (1.544Mbps) link into the Frame-Relay cloud.
R2 is a sales office and they dont need internet anyways. The internet is for closers!!

Anyways, R2 just gets a 256Kbps link. To accurately show this in OSPF we will set the bandwidth on the s0/0 interface of R2. First, lets check the cost value of S0/0 for comparison later.



So a serial interface defaults to 1.544Mbps and OSPF calculates this as a cost of 64.



The cost now jumped up to 390 which is expected. The formula used to calculate the OSPF cost is:



cost= 10000 0000/bandwith in bps
or 
cost= 100,000,000/256,000 = 390.625 (close enough)


You can also see R1 is receiving the LSAs with differing cost (aka metrics) for the link.



Now we can check the routing table to see which route is installed. My money is on R3's route.


CRAP! Still using both routes for load balancing  Notice how the metric stays the same too. Since we cannot influence the cost via the cost or bandwidth command lets change the network type to Multipoint nonbroadcast and see what happens.



There you go. R1 is now choosing the route to R3 over R2 since it has the lower cost or metric.

An alternative to this approach is to not convert the links to Multipoint non broadcast and just add neighbor statement to the OSPF process and specify the cost for each  Essentially it does the same thing since neighbors will now use unicast HELLOs but you do not get auto-discovery of your neighbors.