MOSPF

from Wikipedia, the free encyclopedia

Multicast Open Shortest Path First (MOSPF) can be used in computer networks to process parcels according to the multicast principle. MOSPF is the multicast extension to the intradomain routing protocol OSPF (Open Shortest Path First). MOSPF enables multicast routing within an OSPF domain . Here, however, special attention must be paid to the fact that this domain cannot be / may not be of any size and that the routing calculation works efficiently and quickly.

functionality

In OSPF, each router manages a complete database that describes the entire network topology . By introducing a new type of link state record - the group membership LSA - it is possible with MOSPF to save the location of all multicast participants in a multicast group in the topology database. Since OSPF is a link-state protocol, this information is available on all routers in the domain. Group membership LSAs are generated by those routers that determine group membership on a subnet ( LAN ) via IGMP . Like conventional router and network LSAs, these LSAs only remain within an OSPF area and are therefore not forwarded to other areas or even other autonomous systems .

In combination with the conventional router and network LSAs ( topology database ), starting from the router behind which there is a multicast source (source), an implicit source base tree - a shortest path tree , can be created for each source . engl. Shortest Path Tree (SPT for short), which has the multicast source as its root - calculated by the network. No additional multicast routing protocol is required because the conventional unicast routing of OSPF is also used.

On-demand calculation

Theoretically, the SPT can be calculated via RPF ( Reverse Path Forwarding ) and pruning on the router before a multicast packet is sent (since the entire topology database is available in each router). However, this can lead to performance problems, since the router has to calculate its own SPT for each multicast source. These calculations can also bring routers equipped with comparatively powerful CPUs to their knees.

Instead of this automatic calculation of all SPTs, the SPTs are therefore calculated as required (computation on demand): The SPT is only calculated when the first multicast packet from a multicast source for a specific multicast group arrives. Possible branches in this calculated SPT that do not have any multicast receivers are removed from the tree (pruning). Nevertheless, this method can also lead to a performance problem if there are many different sources behind a router and many multicast receivers log on and off frequently. Such behavior in turn leads to constant recalculations of the SPT.

If an SPT has now been calculated, this information is temporarily stored in a cache on the router and used for later communication between the multicast source and the corresponding multicast group.

Inter-area routing

OSPF makes it possible to subdivide an OSPF domain (e.g. an autonomous system ) into several so-called areas . If this is done, not every router has a complete topology map of the OSPF domain. Instead, each router only knows its area (s), while communication between areas takes place via the ABRs ( Area Border Routers ). While this is advantageous for unicast OSPF; however, with multicast OSPF, full SPTs cannot be calculated. This leads u. This can lead to inefficient multicast routing.

In order for MOSPF to work between two areas, the area-border routers must know about the multicast behavior of the network. Such routers are known as Multicast Area Border Routers (MABR). A MABR will summarize all multicast information and send the corresponding LSAs to Area 0 (Note: Area 0 does not distribute this received information to other areas, ie this procedure can only be used if the multicast sources are in the area 0). If multicast sources but also lie outside the Area 0, so this problem is by setting a wildcard - bit solved each MABR in the router-LSA. This bit means that every MABR becomes a member of every multicast group and thus also sucks in the multicast traffic of a source that is not in area 0. The multicast traffic thus arrives at Area 0 and this can then forward the multicast traffic to the affected areas in accordance with the group memberships known in Area 0. However, this can lead to undesired network traffic if no “real” multicast sources exist.

Alternatives

swell