Solved: Type 2 LSA routing calculation----Help

frecolwang · ‎09-08-2009

Hello,

In this diagram,router-1.1.1.1----router-2.2.2.2 is Ethernet

on the router-1.1.1.1,I want to route to 6.0.0.2/8,I think I use type 1 LSA can calculation the route,I want to know where the type 2 LSA will be use?

Peter Paluch · ‎09-08-2009

Hello,

If you configure that network you will see that the LSA1 of both routers do not indicate a direct connection between the two routers. Rather, each router's LSA1 will contain a link description to the transit network 5.0.0.0/8. As discussed in other threads a few days ago, a link to another object is indicated in LSA1 by referencing that object's LSID (Link State ID) in the LSA1's LID (Link ID). In this particular network, one LID entry in each router's LSA1 would contain the LSID of the LSA2 representing the multiaccess network. The path from the left router to the network 6.0.0.0/8 will be expressed in terms of LSAs as:

R1's LSA1 -> Network 5.0.0.0/8 LSA2 -> R2's LSA1

This is where the LSA2 comes in. Note that the LSA2 that describes the network cannot be created by the network itself - the cables, hubs, access points or switches do not speak OSPF so some member router of the network 5.0.0.0/8 must create the LSA2 on behalf of the network in addition to creating its own LSA1. This responsibility is given to the Designater router on the segment.

But your thinking is not incorrect. In fact, having or not having an LSA2 for a network is the fundamental difference between multiaccess and point-to-multipoint network types in OSPF. You could indeedd indicate a direct connection from one router to another in the LSA1 - that is absolutely correct. However, if you had 50 routers on a segment, each LSA1 would need to contain 49 link descriptions, from one router to each other. Moreover, if you added a new router on the segment, all existing 50 routers would need to reoriginate their LSA1 to include the connection to the new router. This would require having large LSA1 packets and each addition or removal of a router from the network would cause significant flooding of LSA1. On the other hand, this approach gives you the possibility to indicate a different cost to each neighbor on the network and also it solves the problem if the network does not allow direct communication between every two routers (such as hub-and-spoke or partially meshed Frame Relay network). This is how the point-to-multipoint OSPF network type works.

If you have a LSA2 that represents the network, then each router indicates only a connection to the common network in its LSA1 instead of declaring a link to each neighbor. So for any count of routers on a common segment, the LSA1 of these routers contains only a single link to the network. The LSA2, on the other hand, contains the list of all routers on the segment. It's like saying "I am connected to this network and this network connects to everybody else here", so in effect, "through the network, I am connected to everybody else here". If a new router connects to this network, only the LSA2 for the network needs to be reoriginated, not the LSA1 of all member routers. This is how the multiaccess OSPF network type works. However, this approach is not working well for networks like hub-and-spoke or partially meshed Frame Relay because the LSA1+LSA2 in effect create a model of fully meshed neighbors in a network while in reality, they might not be fully meshed. This in turn creates reachability issues that you must solve with additional configuration (such as adding static IP/DLCI mappings).

So each approach has its positive and negative consequences and there is no ideal option. Some networks work better with one approach, some other with the second. It's up to knowledge of the network administrator to choose what is the best for his needs.

Best regards,

Peter

View solution in original post

Peter Paluch · ‎09-08-2009

Hello,

If you configure that network you will see that the LSA1 of both routers do not indicate a direct connection between the two routers. Rather, each router's LSA1 will contain a link description to the transit network 5.0.0.0/8. As discussed in other threads a few days ago, a link to another object is indicated in LSA1 by referencing that object's LSID (Link State ID) in the LSA1's LID (Link ID). In this particular network, one LID entry in each router's LSA1 would contain the LSID of the LSA2 representing the multiaccess network. The path from the left router to the network 6.0.0.0/8 will be expressed in terms of LSAs as:

R1's LSA1 -> Network 5.0.0.0/8 LSA2 -> R2's LSA1

This is where the LSA2 comes in. Note that the LSA2 that describes the network cannot be created by the network itself - the cables, hubs, access points or switches do not speak OSPF so some member router of the network 5.0.0.0/8 must create the LSA2 on behalf of the network in addition to creating its own LSA1. This responsibility is given to the Designater router on the segment.

But your thinking is not incorrect. In fact, having or not having an LSA2 for a network is the fundamental difference between multiaccess and point-to-multipoint network types in OSPF. You could indeedd indicate a direct connection from one router to another in the LSA1 - that is absolutely correct. However, if you had 50 routers on a segment, each LSA1 would need to contain 49 link descriptions, from one router to each other. Moreover, if you added a new router on the segment, all existing 50 routers would need to reoriginate their LSA1 to include the connection to the new router. This would require having large LSA1 packets and each addition or removal of a router from the network would cause significant flooding of LSA1. On the other hand, this approach gives you the possibility to indicate a different cost to each neighbor on the network and also it solves the problem if the network does not allow direct communication between every two routers (such as hub-and-spoke or partially meshed Frame Relay network). This is how the point-to-multipoint OSPF network type works.

If you have a LSA2 that represents the network, then each router indicates only a connection to the common network in its LSA1 instead of declaring a link to each neighbor. So for any count of routers on a common segment, the LSA1 of these routers contains only a single link to the network. The LSA2, on the other hand, contains the list of all routers on the segment. It's like saying "I am connected to this network and this network connects to everybody else here", so in effect, "through the network, I am connected to everybody else here". If a new router connects to this network, only the LSA2 for the network needs to be reoriginated, not the LSA1 of all member routers. This is how the multiaccess OSPF network type works. However, this approach is not working well for networks like hub-and-spoke or partially meshed Frame Relay because the LSA1+LSA2 in effect create a model of fully meshed neighbors in a network while in reality, they might not be fully meshed. This in turn creates reachability issues that you must solve with additional configuration (such as adding static IP/DLCI mappings).

So each approach has its positive and negative consequences and there is no ideal option. Some networks work better with one approach, some other with the second. It's up to knowledge of the network administrator to choose what is the best for his needs.

Best regards,

Peter