I am currently studying OSPF and would be very grateful if anyone could clear up some confusion for me.
I have read that a Designated Router is selected per ethernet segment (i.e per subnet). Now in an ethernet environment say with 4 routers in an area there will likely be a minimum of 4 ethernet segments linking them. In this situation surely that means there are 4 DRs within a single area but there is only supposed to be one DR in an area to save LSA overhead. Must all routers within an area be connected by a single segment? I've read several expalnations and it's still no clearer to me so would really appreciate someone clearing this up!
OSPF elects a DR and BDR on an ethernet network per ethernet segment not per area. So for example if you have 100 ethernet segments within an area you will end up with 100 DR's and BDR's.
DR and BDR save LSA overheads because on any segments there can be multiple OSPF speaking routers. By electing a DR/BDR and only sending LSA updates to the DR which then propogates them you are saving on LSA overhead. You are also more importantly saving on every router on an ethernet segement having to an adjacency with every other router. They only need to have adjacencies with the DR and BDR.
Hope this has helped
Thanks for your reply Jon.
From what you've said then the election of a DR and BDR will only save LSA overhead if there are more than 2 routers on an ethernet segment (a DR, BDR and 2 DRothers) If there are only 2 routers on the segment they will need to form an adjacency anyway so there would be no benefit in electing a DR?
Is the preferred design in ethernet to have lots of routers conected by one ethernet segment?? because (in my experience) I can't envisage a situation where this is practical.
The material on OSFP does suggest only one DR and BDR per area hence the confusion!
Okay imagine a pair of 4500/6500 switches running HSRP/GLBP. So you have 2 switches each with a supervisor card. You create a vlan eg vlan 10 and on each switch create a vlan 10 interface. So now you have 2 routed interfaces running OSPF in the same ethernet segment.
You may well then have standalone routers/firewalls that run OSPF that you allocate into the same ethernet segment. It's quite easy to actually end up with an ethernet segment with quite a few more than 2 OSPF speaking routers.
Thanks - that's helped - my experence is primarily with EIGRP so haven't come across this kind of setup.
Can I ask one more question. In a NBMA Frame Relay Network why doesn't multicast work? I have to admit I know very little about FR and my understanding of it is that it encapsulates IP within a Layer 2 frame. Once this frame is stripped off at the remote end you're left with a muticast IP frame??
Thanks for your patience!
Multicast works quite well over NBMA Frame-Relay.
You need to make sure to include the broadcast keyword on the mappings.
ip address x.x.x.x y.y.y.y
frame-relay map ip x.x.x.x dlci yyy broadcast
I would like to add to what Edison says about multicast. Well, actually about Frame Relay.
First off, OSPF. OSPF works fine in frame-ralay but you have to be careful about the design. If you have a hub-and-spoke topology (which is usual for frame-relay), then you MUST arrange for the hub router to be the DR, basically because you cannot form adjacencies spoke-to-spoke.
Now, for multicast. Multicast does work fine in frame relay environments, but the same sort of design constraints can apply. In particular, if you want to pass traffic spoke-to-spoke with sparse-mode multicast, you should add ip pim nbma mode on the hub router. The subject is quite complex, and you would do well to read a lot about it before implementing.
To avoid the DR election on an Ethernet link between two routers only, f.e. GE back to back between two routers, you can use a little trick.
Change the OSPF network type from BROADCAST (default for ethernet, needs DR/BDR) to POINT-TO-POINT (no DR/BDR required).
ip address 10.1.1.1 255.255.255.252
ip ospf network-type point-to-point
This way you get the desired adjacency without the DR/BDR election.
You are right, there will be one DR per ethernet segment. When there are many segments within an area then so many DR's as segments. In such a practical case where there are many segments on a single ethernet, one can allocate a separate subnet for all devices participating in ospf process & thus reduce the overhead of DR / segment. I've seen this in many scenario's.
hope that clarifies.
pls rate all helpful posts.