DR & BDR Election on Serial Link.

[Ravi_916]

Could You Tell me the Actual Reason,  Why DR & BDR Election not happens on Serial Link.

i know its a Point-to-Point Network, election is not required but something actual reason behind of that.

[Leonid Voronkin]

DR & BDR elects in broadcast domain to reduce quantity of LSAs. And it uses additional LSA Type 2 for it. You don't need this mechanism when the link has only two routers. Because it will only increase LSA quantity

[Georg Pauwen]

Hello,

in addition to Leonid's post, I can imagine the theory behind it being that serial links are (very) low speed by nature, so you want to reduce any unncecessary routing protocol traffic as much as possible.

[Cristian Matei]

Hi,

     Based on the OSPF network-type, which you can manually change on each link individually, the OSPF routers will silently agree (there is no actual negotiation) if there is DR/BDR election or not; this silent negotiation is very important, they need to run same network type (all routers on a segment need to run a network type that has DR/BDR election, or a network type which doesn't have DR/BDR election; for example running point-to-point on one side and point-to-multipoint on the other side will perfectly work, as both network types do not run DR/BDR election), as based on that, the graph of nodes and the SPF tree is build differently.

    The only reason for the existence of DR on broadcast and non-broadcast network types, is due to the inherent specifics of these networks, where multiple devices share the same segment, which mixed with the LSA flooding concept of OSPF, could cause chaos and segments where convergence may not work as intended; so they had to optimize and scale this, which means they invented this role of DR (they actually tried to copy the DIS role from ISIS, but the end result is not as good as in ISIS), which is the guy in charge with LSA flooding on those links; everyone sends their LSA's to the DR (all routers build full adjacency, so actuary send LSA's, only with DR and BDR) and the DR sends it further. You can look at the DR, like you look at an BGP route-reflector, the BGP route reflector reflects routes, by default without touching them, the DR reflects LSA's. The BDR, which does not exist in ISIS, doesn't do anything, it just sits and waits for the DR to be gone, so it can become the DR itself. Why do we have a BDR in OSPF and not a BDIS in ISIS?  Because otherwise, per OSPF design, if the DR fails and there is no BDR, you have network down event. In IS-IS, on any given link, you'll have a full mesh of adjacencies (full) and routers exchange everything directly, the DIS only makes sure that everyone has the same vision, through periodic updates. With OSPF, routers build full adjacency only with DR, with regular routers they remain in two-way, so if the DR is gone and there would be no BDR, all routers in that area, will have to build full OSPF adjacency with a newly negotiated DR, which means everyone in that area would have to rebuild the SPF tree, which is network down event. For this reason we have the BDR, and all routers have a full adjacency with BDR as well, so when it gets promoted as DR, network impact is not big.

       From my point of view, ISIS is way simpler, way more scalable and way better designed than OSPF; OSPF is just A bad copy of ISIS, the only advantage being is that is more feature rich, but the arhitecture itself is not really good. As a matter of fact, even today, OSPF has two corner cases where it fails to create a loop free environment.

Regards,

Cristian Matei.

[Joseph W. Doherty]

If, as you note, such an election isn't required, why would there be a reason beyond that? If fact, on Ethernet links that are actually p2p, but "seen" by OSPF as a shared media, i.e. not p2p, it's good to define the link's network type as p2p to avoid the unnecessary overhead (as described by the other posters; also not just link bandwidth by router CPU) that goes along with DR/BDR.

[Giuseppe Larosa]

Hello @Ravi_916 ,

>> i know its a Point-to-Point Network, election is not required but something actual reason behind of that.

only two routers so network type point to point to be honest OSPF is used in this way in modern fiber based point to point links, that is avoiding  for the DR/BDR election to occur as it does not provide any benefit.

And OSPF is used in some of the largest european service providers with just using

int gi0/0

ip ospf network point-to-point

That's all.

Be aware that ALL DR routers is 224.0.0.6 and ALL OSPF routers is 224.0.0.5 when a router sends an update over a shared segment it sends to all DR routers 224.0.0.6 only the DR and BDR listern to it. The DR receives the update and then sends out the new OSPF update that contain one or more LSAs ( data structures) to 224.0.0.5 the original sender will simply ignore the update all the DR other routers will receive and process it and send an LSA ack back to DR.

the DR/BDR minimizes the number of OSPF full adjacencies to be built and the number of Exchange/full link state DB exchange on a shared segment instead of N*(N-1)/2 (a full mesh) they are + 2*(N-2) +1 (DR to BDR).

In the past there were even 50 OSPF routers in the same LAN but the OSPF RID and priority come into play only when no DR is present. the OSPF DR and BDR cannot be pre-emepted if a new router with better OSPF RID or OSPF priority was connected to the shared segment it could not become DR immediately.

As a result of this in real world the OSPF DR is usually the device with the greatest uptime rather then the device with the highest OSPF pri and OSPF RID.

It happened in one case that the DR was a SUN workstation.

There is parameter called wait time equal to 40 seconds that tells how long a router waits before declaring itself the DR for the segment

In exam questions usually you can read something like:  the switch is reloaded who between R1-R6 will become the DR ?

Hope to help

Giuseppe