At 9:38 PM +0000 7/8/03, Priscilla Oppenheimer wrote: >hebn9999 wrote: >> >> hello,everyone: > >> OSPF use raw socket (datagram) to communicate with peers. In > > general, layer 2 frame has a MTU of 1500 bytes.
I'm not sure I'd call it a strict datagram protocol. In some cases, it's acknowledged datagram, either directly or indirectly acknowledged. In other cases, such as database initialization, there's clearly a connection -- just not a TCP connection. > > how does cisco router propagate router-lsa whose size exceed >> 1500 bytes(more than 122 links in one area)? > >Well, I don't have a definite answer, but I'll discuss it with you in the >hopes of lighting a fire under one of the OSPF experts on this list. Howard? >Chuck? Peter? Where's Pamela when we need her? :-) It's fairly simple in Wellfleet/Bay/Nortel code -- you send multiple LSAs. There's nothing that says all the database entries of one type have to fit in the same LSupdate. Certainly, in a large area, there will need to be more than one LSupdate to convey all the Type 2 LSAs. > >OSPF runs directly above IP. I don't know if that could be called "raw >socket" which is a UNIX thing? My perception is that with Cisco IOS, OSPF >calls IP with a set of parameters and lets IP handle the rest. So maybe >that's sort of raw. I'd have to go look at Moy's or another UNIX based implementation to see how the calling is done. IOS is not UNIX based. Several other vendor implementations run a realtime OS such as VXworks. > >I can say this: The OSPF packets I have seen coming out of Cisco routers >have the IP fragmentation bit set to "May Fragment." This makes me think >that Cisco's OSPF relies on IP to push the bytes into the data-link-layer >frame and fragment if necessary. > >The OSPF RFC (RFC 2178) says this: > >"OSPF does not define a way to fragment its protocol packets, and depends on >IP fragmentation when transmitting packets larger than the network MTU. If >necessary, the length of OSPF packets can be up to 65,535 bytes (including >the IP header). The OSPF packet types that are likely to be large (Database >Description Packets, Link State Request, Link State Update, and Link State >Acknowledgment packets) can usually be split into several separate protocol >packets, without loss of functionality. This is recommended; IP >fragmentation should be avoided whenever possible." > >Unfortunately, that's not very clear. It implies that the recommended method >is for OSPF to split its own protocol packets. But that the method for doing >this is undefined and that's OK because OSPF can depend on IP to do >fragmentation. Fragmentation gets scary when you are doing real-time control plane traffic. > >Cisco routers tell each other their MTU in database description packets, per >RFC 2178. Until recently, if the routers didn't agree on the MTU, they >wouldn't become adjacent. A recent IOS version supports telling a router to >ignore the other side's MTU so they can still become adjacent. > >That doesn't answer your question, but maybe there are some hints in the >article that discusse the "ip ospf mtu-ignore" feature here: > >http://www.cisco.com/warp/public/104/12.html Message Posted at: http://www.groupstudy.com/form/read.php?f=7&i=72044&t=72024 -------------------------------------------------- FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
