Link state protocols - Hello & dead timer

[Ratheesh mv]

As we know that for forming OSPF neigborship hello & dead timer must match but if we come to ISIS doesn't demand to match hello and dead timer. Even though both are link state routing protocol their parameters are completely diff to form neigborship in hello and dead timer point of view. I would like to know the exact pinpoint of behind matching hello and dead timer in OSPF protocol.(In advanced level) .If OSPF also operates without matching hello and dead timer what is the issue ?Why did we design OSPF like this ?

Thanks in advance 

[Joseph W. Doherty]

Cannot say for sure, unless RFC details reasoning behind that OSPF design decision.  But if I were to hazard a guess, OSPF might require the match as a simple form of a sanity check, i.e. the two neighbors are really intended to be neighbors.  And/or to insure both neighbors really agree on suitable hello and dead timers.  (For the latter, although not as important on modern equipment, what if one more powerful L3 router "forces" a slower neighbor to use a more CPU intensive hello and dead timer.  And/or again for much earlier WAN links, which might be very slow [i.e. less than 64 Kbps], how much bandwidth will "fast" timers cause, again, if "forced" by only one neighbor.)

[Richard Burts]

The original poster seems to believe that Hello and Dead timers are part of link state protocols. I do not believe that this is the case. They are used in OSPF which is a link state protocol. But as the original post points they are not used in ISIS which is another link state protocol.

Like Joseph I have not searched the RFCs for OSPF to see what they say about Hello and Dead timers. But I believe that there is a reasonable explanation. As Joseph suggests one aspect may be as a sort of sanity check to make sure that the neighbors are really compatible. But I believe that another aspect is that these timers function to make sure that the neighbor is still functioning and that LSAs received from that neighbor are still valid.

[Martin L]

Best book for such deductive questions (why/why not) for OSPF is by its developer J. Moy, "ospf-Anatomy of an Internet Routing".  The author walks from OSPF v1 to v2 and explains why this and that was necessary.  In case of timers, there is no great explanation but the author mentions the use of Hellos to discover neighbors on NBMA segments.

the author also says "Hello packets also serve as a method of last resort to detect when a neighboring router has failed, although usually the failure is detected much more quickly by the data-link protocol (for example, PPP's link-quality monitoring). When no Hello packet has been received from the neighbor for ospf RtrDeadlnterval, the neighbor is declared in operational. To ensure that a couple of lost Hello packets do not incorrectly cause a neighbor to be declared in operational, ospf RtrDeadlnterval should always be several times the ospf if Hellolnterval."  The story behind "waiting timer" is more interesting and significant (also explained in the book.)

Regards, ML
**Please Rate All Helpful Responses **

[Martin L · Accepted solution]

Regarding ISIS, similarly to OSPF, purpose of Hellos is discovery IS-IS neighbors on a link and to form adjacency.  True, there is no Dead timer but there is Hold timer.  This hold timer informs its neighbors how long they should wait before declaring the router dead (usually 3x hello interval).  Just like in EIGRP, ISIS Hellos between two neighbors do not have to match; each router honors the hold time advertised by its neighbor.

ISIS has less network types than OSPF; therefore, ISIS is less strict when comes to establishing adjacency. ISIS Hello PDU focus more on its capabilities.  Note that you can mix certain OSPF types of networks as long as timers are matching.  So, I think in order to extend number of various network types (especially on NBMA) and allow to "mix of them", OSPF engineers add strict matching of Hello and Dead timers requirement to keep adjacency.

Regards, ML
**Please Rate All Helpful Responses **