EIGRP route summarisation

nel

OK so ive started hitting the bsci self study book after having a few weeks off from finishing the ccna and have my first question

OK, so i know what this is and why its used etc but in the Teare self study book it mentions the hub and spoke design as the most commonly used layout but in the few examples used in the book it places the point of summarisation in different points of the network.

For example on page 118 fig 3-28 it displays a large scalable network with the summarisation being injected into the core and the summarisation on the border routers (or im unsure if this is a core router?) but on page 126 fig 3-32 it shows a hub and spoke setup with the summarisation taking place at the point of the core.

Am i correct in thinking the summarisation should take place at the core which is then placed in the spoke's routing table as a summary and inject a default route into the stub routers so any queries outside the summary range are passed to the core? Or is it best to advertise at the edge routers so the core has a more efficient routing table?

Or am i wondering off track and into the design exams?

Thanks

Plazma

Generally from my experience you configure the EIGRP stubs and you don't need to summarize.. EIGRP allows you to summarize on the interface level.. so literally you could do it anywhere. It is best and logical practice to summarize at logical boundaries.. which can vary greatly depending on your network scheme.

nel

Hi,

i know that it is done at the interface level but i guess im trying to say what are the factors for choosing a specific point? are there best practice design guides for this topic?

So say if what teare says is correct and hub and spoke are most commonly used where do most of you guys summarise in the real world?

Its slightly irrating as i know what it does etc its just i want to know the full picture and so i can practice it in my labs and the book seems to assume you would know where.

So for example say you had a network in europe and you had your core then you split it into geographical region...say UK, france, italy, spain. Am i right in thinking you would have an edge router in each location which would then summarise its address scheme to the core?

kpjungle

nel wrote:

Hi,

i know that it is done at the interface level but i guess im trying to say what are the factors for choosing a specific point? are there best practice design guides for this topic?

So say if what teare says is correct and hub and spoke are most commonly used where do most of you guys summarise in the real world?

Its slightly irrating as i know what it does etc its just i want to know the full picture and so i can practice it in my labs and the book seems to assume you would know where.

So for example say you had a network in europe and you had your core then you split it into geographical region...say UK, france, italy, spain. Am i right in thinking you would have an edge router in each location which would then summarise its address scheme to the core?

It really depends on the size of the network as well. If we take your example, and you have a HQ which is the "hub" of your network, and you have spokes in different countries, you could pass a default route onto the spokes. Or configure them as eigrp stubs. Either way, if you pass a default route or the stub you have made sure you have your queries under control. If the remote site has subnetworks, and you ofcourse have made it hierarchical , then you can summarize a route to the core, which will reduce the amount of individual routes the core network has in its routing tables.

As always, the goal is to optimize the network, make the routing tables small enough, and in EIGRP networks, bounding the queries.

nel

Thanks for clearing that up guys!

tech-airman

nel,

nel wrote:

OK so ive started hitting the bsci self study book after having a few weeks off from finishing the ccna and have my first question

OK, so i know what this is and why its used etc but in the Teare self study book it mentions the hub and spoke design as the most commonly used layout but in the few examples used in the book it places the point of summarisation in different points of the network.

For example on page 118 fig 3-28 it displays a large scalable network with the summarisation being injected into the core and the summarisation on the border routers (or im unsure if this is a core router?) but on page 126 fig 3-32 it shows a hub and spoke setup with the summarisation taking place at the point of the core.

The two routers Router A and Router B would be in the WAN Module of the Enterprise Edge Functional Area and not the Core module of the Enterprise Campus functional area.

nel wrote:

Am i correct in thinking the summarisation should take place at the core which is then placed in the spoke's routing table as a summary and inject a default route into the stub routers so any queries outside the summary range are passed to the core? Or is it best to advertise at the edge routers so the core has a more efficient routing table?

As previously explained, your use of the term "core" is incorrect. The system that you sorta need to understand is the "hub and spoke of hub and spokes." It may also be known as an "Extended Star" topology. What I mean is the following...

                                    +---z---(Branch1)
                                   /
                  +---z---(Region1)-------z---------(Branch2)
                /                  \
              /                     +---z---(Branch3)
(HeadQuarters)
              \                     +---z---(Branch4)
                \                  /
                  +---z---(Region2)-------z---------(Branch5)
                                   \
                                     +---z---(Branch6)

So the Region1 router is a hub for the spokes to Branch1, Branch2, and Branch3. The Region2 router is the hub for the spokes to Branch4, Branch5, and Branch6. However, Region1 and Region2 are themselves spoke routers for the HeadQuarters hub router.

Let's say here's the network assignments for the LAN side of the Branch routers:

1. Branch1 = 192.168.1.0/24
2. Branch2 = 192.168.2.0/24
3. Branch3 = 192.168.3.0/24
4. Branch4 = 192.168.4.0/24
5. Branch5 = 192.168.5.0/24
6. Branch6 = 192.168.6.0/24

Now, without summarization, the Region1 router has to carry in it's routing table the following networks:

1. Region1 routing table
   [list=1:df42007253]
2. Branch1 = 192.168.1.0/24
3. Branch2 = 192.168.2.0/24
4. Branch3 = 192.168.3.0/24

[/list:o:df42007253]

The Region2 router has to carry in it's routing table the following networks:

1. Region2 routing table
   [list=1:df42007253]
2. Branch4 = 192.168.4.0/24
3. Branch5 = 192.168.5.0/24
4. Branch6 = 192.168.6.0/24

[/list:o:df42007253]

That means, HeadQuarters router should be carrying the following networks in it's routing table:

1. HeadQuarters routing table
   [list=1:df42007253]
2. Branch1 = 192.168.1.0/24
3. Branch2 = 192.168.2.0/24
4. Branch3 = 192.168.3.0/24
5. Branch4 = 192.168.4.0/24
6. Branch5 = 192.168.5.0/24
7. Branch6 = 192.168.6.0/24

[/list:o:df42007253]

So the best points of summarization would be the Region1 and Region2 routers. So for Region1, the appropriate manual summarization network address would be 192.168.0.0/23. For Region2, the appropriate manual summarization network address would be 192.168.4.0/23. The effect of implementing these manually summarized network addresses will be most significant on the HeadQuarter router's routing table which would go from:

1. HeadQuarters routing table
   [list=1:df42007253]
2. Branch1 = 192.168.1.0/24
3. Branch2 = 192.168.2.0/24
4. Branch3 = 192.168.3.0/24
5. Branch4 = 192.168.4.0/24
6. Branch5 = 192.168.5.0/24
7. Branch6 = 192.168.6.0/24

[/list:o:df42007253]

to...

1. HeadQuarters routing table
   [list=1:df42007253]
2. 192.168.0.0/23
3. 192.168.4.0/23

[/list:o:df42007253]

So there's a routing table size reduction on upstream routers from the perspective of the routers where manual summarization was configured. Another benefit of manually summarized networks is that if any of the Branch routers goes down, it will not disturb the converged state of the network between the HeadQuarters and Region routers.

So to directly answer your question of "Am i correct in thinking the summarisation should take place at the core which is then placed in the spoke's routing table as a summary and inject a default route into the stub routers so any queries outside the summary range are passed to the core?" no you are not correct in thinking that the summarization should take place at the core. Summarization should occur beFORE you reach the core, meaning summarization should occur in the Distribution layer. As explained above, the hub in the Core Layer leads to spokes in the Distribution Layer. But the spokes in the Distribution Layer are hubs themselves for the spokes in the Access Layer. So when you mention "....which is then placed in the spoke's routing table as a summary and inject a default route into the stub routers...." seems to suggest that you are confused about the direction of summarization. Routes are not pushed from the Core to the Distribution to the Access layers but the Access routers teach the Distribution routers who then learns. Then the Distribution routers turn around then teach the routes to the Core routers who learn. When you mention ".... so any queries outside the summary range are passed to the core..." is itself confusing. It is when a Core router is somehow notified of a network outage when the Core router sends out a Query packet to it's EIGRP neighbor routers, in this case, the Distribution routers. The purpose of the Query packets from the Core EIGRP routers to the EIGRP neighbor routers is to find out if anyone else knows how to get to the network that was just reported down. However, the Core EIGRP routers were notified that a network went down because there was an ABSENCE of manual summarization. Now, if manual summarization was configured on the Distributon routers, then the Distribution routers would NOT send a "network down" message to the Core routers thus avoiding the Query and Reply process which may result in a "stuck in active" or SIA condition. So when you say "... is it best to advertise at the edge routers so the core has a more efficient routing table..." it depends if you mean "edge routers" as the Distribution routers and not the Access routers.

nel wrote:

Or am i wondering off track and into the design exams?

Thanks

Since the 640-801 BSCI exam is a required exam for the CCDP certificate, I beleive that it is important for you to start to learn how the network designers think so you'll understand as a network implementor why you're configuring what you're being told to configure.

nel

Woah, tech-airman. Great post.

Now the way you have just described it there has made perfect sense to me and i felt you linked it with the design very well for that example - better than the book infact. I really do appreciate your effort for that post.

techman wrote:

Since the 640-801 BSCI exam is a required exam for the CCDP certificate, I beleive that it is important for you to start to learn how the network designers think so you'll understand as a network implementor why you're configuring what you're being told to configure.

I totally agree with that statement. Maybe its time to hit the design aspects a little more before i go any further?