how routing decision was made with route aggregation?

[Steve Zhou]

Hi,

As we all know, route aggregation reduce the size of routing table and save processing power when re-calculating routes on the main/core router. Could you help me to understand how does the main/core or upper stream router make the routing decision in the following figure?

The route entry 172.24.0.0/13 summarizes 8 subnets (172.24.0.0/16 ~ 172.31.0.0/16) here, how does the upper stream router know where to send the packet destined to 172.26.0.5 if it has only one route entry (the summarization) in its route table? Did I miss anything here? Thank you!

[curdubanbogdan]

Before all this what routing protocol are you using?

[Giuseppe Larosa · Accepted solution]

Hello @Steve Zhou ,

the summary route 172.24.0.0/13 is sent downstream to routers that are more far from the specific aggregated subnets.

In addition to this, the summarizing node does not lose the knowledge of the component routes.

So when a packet with destination 172.26.0.5 arrives at the node that performed summarization the longest prefix match is found and this provides the outgoing interface.

Being 172.26.0.5 an host directly connected to an interface of the router the adjacency CEF table will provide all the info to perform packet rewrite (either because an ARP entry already exists for the host or by performing an ARP request and processing the ARP reply).

The idea behind route summarization is that for all networks devices on the left there is no need to know the details of the component routes to be able to correctly route a packet within 172.24.0.0/13 aggregate route.

Think of the summary route as only sent out the interface on the left but not used by the aggregating router that knows the component routes.

Most routing protocols like EIGRP or OSPF install also a discard route to null0 with an high AD to be able to deal with the case of a failed component route to avoid possible routing loops.

There are great differences on where you can perform route summarization that are protocol dependent.

OSPF can do it only at area border, EIGRP allows to do it where you want but you need to take care of the admin distance of the generated summary route to avoid possible blackholing.  EIGRP summary route have a default AD of 5 that can create issues.

Hope to help

Giuseppe

[Richard Burts]

@Steve Zhou @Giuseppe Larosa has provided a good explanation. I would suggest looking at the question from a slightly different perspective. I would start by making sure that I correctly understand your diagram and your question. There are several downstream routers where individual subnets are connected which are connected to a middle router which will do aggregation. The middle aggregating router is connected to the main core router. The main core router has a single route for 172.24.0.0/13 with the aggregating router as the next hop. Your question is how does the main core router know what to do if it is trying to forward a packet to 172.26.0.5. If any of this is not correct then please provide clarification.

In this case the forwarding logic for the main core router is quite simple. It has an entry in the routing table with the aggregating router as the next hop. The main core router does not need to know exactly where 172.26.0.5 is connected. It only needs to know what is the next hop to get to that destination. Note that this situation is the same no matter what the routing protocol (EIGRP, or OSPF, or BGP etc) and in fact would be the same even if the network were using static routes. The mechanics of how to do the aggregation would be different, but the fundamental forwarding logic is the same.

[Steve Zhou]

Hi  @Richard Burts ,

I am in learning the theory of route aggregation. With the help from @Giuseppe Larosa  and you, I think I've cleared my question. As you mentioned, the core router (which I didn't draw) is the router who was benefit from the aggregating route because it only needs to know where to get to 172.24.0.0/13 without the details to each of the subnets, which saves its routing table space. On the router who does the aggregation, it still has the routes to each of subnets. I was confused it only has one aggregation route which was a misunderstanding.

Thank you!

Steve Zhou

[Richard Burts]

Steve

Thanks for the update. I am glad that you have a better understanding about aggregation and that our explanations were helpful. This community is an excellent place to ask questions and to learn about networking. I hope to see you continue to be active in the community.

[Joseph W. Doherty]

BTW, it may not be just the (not drawn) core router that would benefit from the aggregate address, it also benefits any other router that's passed the aggregate vs. all the subordinate encompassed subordinate routes.

Also the benefits extend beyond just saving routing table space. For example, you don't need to "slosh" the additional routes through your network nor might you need deal with computing "best path" for those additional routes. You might also avoid some additional route computations when leaf network join or leave a network.

Perhaps, a common example of (an extreme?) aggregation that you might have used, would be usage of a default route. If you've used a default route, you may have only use a statically define version, but some routing protocols also support passing a default route dynamically.

For example, consider if all the routers in your drawing were passed (from their upstream router) only a default route. This might be done in addition to passing, where possible, aggregate route from the downstream router to its upstream router. Also note, doing such aggregates works great for hierarchical topologies, not so great, if at all, for other topologies.

Lastly, on the merits of aggregation, consider what a network without using it might have to deal with if supporting Internet-of-things (using IPv6). (Yea, we're not like those "dummies" that created IPv4 using a 32-bit address, we're all set for the future with a 128-bit address. Ah, let's see how well in works over the next 40 years, as did IPv4; laugh.)

[Steve Zhou]

thanks @Giuseppe Larosa ,

I think I got your point. The upper stream router in the white-boarding also contains the component routes, not just the aggregate route. The aggregate route will be distributed to the left of the world so everyone now knows how to find the network 172.24.0.0/13 without the knowledge subnets within it, that's the job of the aggregating router here who knows the details. 

So it actually saves the routing table space of the routers on the left of the world, not on the aggregating router. My previous thoughts was wrong. Hopefully now I get it clear from you. Appreciate that!!

Regards,

Steve Zhou