non-full routing aggregation problem

[Steve Zhou]

Hi,

When I was calculating aggregating routes for (192.168.0.0/24 ~ 192.168.7.0/24), a quick question jumped into my head was what about 192.168.8.0/24? It is a valid subnet, but it's not part of 192.168.0.0/21. What if I did't have the next 8 subnets block (192.168.8.0/21) defined? Then what would 192.168.8.0/24 be handled in this case? Does it mean that the upper stream router who holds the aggregating route has to advertise it to rest of the world? I know in a real world deployment, maybe no one would do it, I just want to understand the theory behind it.

Thanks!

[Steve Zhou]

Also, it reminds me a general rule of route selection that is the longest prefix match. In a case where on the upstream router, there is an entry for the next /21 subnets block that is 192.168.8.0/21 and another entry for 192.168.8.0/24. For an incoming packet destined to say 192.168.8.100, the down stream router with 192.168.8.0/21 (not in the drawing) will be chosen rather than the one hosting 192.168.8.0/24 subnet, am I right?

[Giuseppe Larosa]

Hello @Steve Zhou ,

>> For an incoming packet destined to say 192.168.8.100, the down stream router with 192.168.8.0/21 (not in the drawing) will be chosen rather than the one hosting 192.168.8.0/24 subnet, am I right? 

No, because the longest prefix match means the most specific prefix first so 192.168.8.0/24 is preferred over 192.168.8.0/21 when the destination is 192.168.8.100. If the destination was 192.168.9.100 then the only matching prefix would be the aggregate 192.168.8.0/21.

About the fact of creating a summary / aggregate route when not all the component routes exist in this moment. This is something that can be done with an appropriate address plan = the missing component routes must not be in use in another part of the network.  The addressing has to be strictly hierarchical and the missing subnets might be created in the mid term in the same site.

Protocols like EIGRP and OSPF provide a discard route feature that allows to handle packets for a non existing subnet.

Hope to help

Giuseppe

[Steve Zhou]

Oh...yes, I made a typo, 192.168.8.0/24 is the one with the longest prefix matching. And if the packets was destined to 192.168.9.100, then we should get a destination unreachable (no route) error if I was correct.

[Richard Burts]

Aggregating to a /21 would include 192.168.0.0 through 192.168.7.255.  If you do not have an aggregation configured for 192.168.8.0/21 then there are a couple of things to consider:

- do you have individual /24 networks in that range?

- If you do have individual networks in that range are they directly connected? In that case you would route directly to them.

- If you do have individual networks in that range do you have a default route (might be a static default route or might be a default route learned from a dynamic routing protocol). It that case you would forward traffic to those networks using the default route.

- If those networks do exist in the network and are not directly connected and do not have a default route then those networks are not reachable (and you have a broken network design).

- you have not provided information about the relationship between the router doing the aggregation and the upstream router. Are both routers using static routing? If so then the upstream router will need to do its own aggregation. Are both routers running some dynamic routing protocol? In that case your router would advertise the aggregate route for the /21 and would advertise the individual /24s that are not aggregated.

Similar considerations come into play for the upstream router. Does it have routing information for the aggregated networks? Does it have directly connected networks in the range being discussed Does it have dynamic learned route (for aggregate and/or individual networks)?

I am not sure that I understand your follow up question about the upstream router and the 192.1468.8.0/24 network. Based on the drawing I am assuming that your aggregating router does have a route for 192.1468.8.0/24 (not clear whether this route would be static or dynamic routing protocol). And I am assuming that the upstream router has a routing entry for 192.168.0.0/21 and a separate routing entry for 192.168.8.0/24. (again not clear whether those routing entries are static or dynamic - and that really does not matter)  The routing logic on the upstream router would use the principle of longest match and choose the single routing entry it has for 192.168.8.0/24. We know that ultimately the packet needs to be sent to the router on the right side of the drawing. But the routing decision on the upstream router does not worry about where it goes ultimately but only decides what is the next hop, which would be your aggregating router.

[Steve Zhou]

The drawing can be misleading. I think I can explain my own question as: since 192.168.8.0/24 cannot be aggregated, it is simply that we won't have an aggregating route on the upstream router (the left one), there will still be a single route to 192.168.8/0/24 on it though. The left-router will also advertise the reachability info to 192.168.0.0/21 and 192.168.8.0/24 to the rest of the world.

I am not considering any specific routing protocol in this case. :)

[Richard Burts]

Thanks for the update. Your explanation here is correct. You can have an aggregate route for the /21 and a route for the individual /24. Note that aggregation can be done with most of the dynamic routing protocols and can also be done with static routing.

Another comment is that aggregation and summarization are pretty much the same thing. As you read about routing protocols you are more likely to see it referred to as summarization. But it is pretty much the same thing.