Добавил:
Опубликованный материал нарушает ваши авторские права? Сообщите нам.
Вуз: Предмет: Файл:
Internet Routing Architectures Second Edition - Cisco press.pdf
Скачиваний:
99
Добавлен:
24.05.2014
Размер:
4.91 Mб
Скачать

Internet Routing Architectures, Second Edition

Forming the Aggregate Based on a Subset of the More-Specific Routes

You have seen that with AS_SET the aggregate will contain a set of all attributes (including AS numbers) that exist in the individual routes being summarized. If the aggregate is summarizing routes that come from different ASs, it becomes useful to specify which routes are being included in forming the aggregate. This would help in a hub and spoke situation in which each of the leaf ASs contains a separate subset of the aggregate that is originated by the hub. When forming the aggregate, the hub AS would exclude the more-specific routes that belong to the leaf AS that needs to receive the aggregate. The aggregate received by the leaf AS would not contain the AS number of the leaf AS, so it would not be discarded. Figure 6- 33 gives an example of where this could be used.

Figure 6-33. Forming the Aggregate Based on a Subset of More-Specific Routes

AS3 is a hub AS receiving routes 192.68.11.0/24 and 192.68.10.0/24 from the leaf ASs AS1 and AS2. Prefix 192.68.11.0/24 has an AS_PATH of 1, and 192.68.10.0/24 has an AS_PATH of 2. When the AS_SET aggregate is being formed by AS3 based on all the more-specific routes, the AS_PATH information would be {1 2}. The aggregate itself, if sent back to either AS1 or AS2, would be discarded for loop prevention. AS1 would see its AS number in the AS_PATH information and would drop the update; the same is true for AS2. If you could specify which more-specific routes can form the aggregate, you could, for example, specify that the aggregate is to be formed based on 192.68.11.0/24 only. This way, the AS_PATH information would be 1 and would not contain AS2. The aggregate could now be sent back to AS2 with no problem. AS2 can use this aggregate to forward traffic to all destinations in AS1.

TIP

See the section "Forming the Aggregate Based on a Subset of More Specific Routes" in Chapter 11.

Looking Ahead

Having mastered the basics of routing protocols and examined the particular attributes of BGP, you are now in a position to begin applying these tools to specific internetworking

page 179

Internet Routing Architectures, Second Edition

topologies. In doing so, you will be juggling a number of overarching design goals— redundancy, symmetry, and load balancing—that are of varying importance depending on a particular network's needs. Sometimes these needs conflict with one another. The meaning of these design goals is discussed in more detail in the next chapter. The attributes covered in this chapter are used in the following chapter to achieve the desired routing design goals.

Tuning BGP capabilities to satisfy a network's needs involves looking both outside and inside the AS. In other words, the policies set by the networks to which you are interconnected, although usually outside your direct control, have practical implications for how you configure BGP at your end. All this becomes clearer in the next chapter, which covers multiple redundancy, symmetry, and balancing scenarios by presenting architectures commonly used on the current Internet.

Frequently Asked Questions

Q—

If my IBGP peers are not directly connected, do I have to use EBGP multihop?

A—

No. There is no restriction on IBGP peers to be connected. EBGP multihop is for EBGP only.

Q—

Should I inject my BGP routes into the IGP for synchronization to take effect?

A—

No. Injecting the BGP into your IGP is not recommended. You should turn the BGP synchronization off. Make sure that this will not result in reachability problems inside your AS.

Q—

Will listing my IGP routes via the network command rather than redistributing the IGP into BGP give my BGP routes more stability?

A—

No. In both methods, the fluctuation of your IGP routes will translate into fluctuation in your BGP routes. The network command only gives you better control and less worry about what your IGP might distribute into BGP if you use redistribution. Combine it with static routes (with distance 254, for example) to Null0 if you want to prevent route fluctuation even if your IGP routing is not stable.

Q—

Do I have to list my connected interfaces by using the network command?

page 180

Internet Routing Architectures, Second Edition

A—

If you want the directly connected subnets to be reachable via BGP, you can advertise them. If not, you don't have to.

Q—

I have two border routers talking EBGP to my provider and IBGP internally. If I list my IGP routes via the network command on both routers, would that create a loop on the IBGP session?

A—

No, you will not create a loop. Actually, doing so gives you more redundancy. If one of your border routers fails, the other border router will still announce the same networks.

Q—

I need to receive only a few routes from my neighbor. Can I filter on my side of the link?

A—

Yes, you can. However, you should ask your neighbor to send you only the routes you need to minimize unnecessary link bandwidth usage and unnecessary route fluctuations. At the same time, use filtering on your end to protect yourself from potential accidents in which your neighbor sends you more routes than you are expecting.

Q—

My provider needs me to send him different local preference on different links. Is that possible?

A—

No. Local preference is defined inside the AS and is not carried on EBGP sessions. However, a more optimal solution might be to ask your provider to set up a route map that does a match based on BGP community strings (such as those defined in RFC 1998), on his side of the peering session, to set the local preference for your routes in his AS. You will have to tag your routes with the appropriate community string in order for the local preference to be set correctly by his peer session. You could also use MEDs for this.

Q—

I am receiving a MED from my provider that is influencing my traffic in a way that conflicts with my IGP. What should I do?

A—

If receiving MEDs is causing you problems, call your provider and ask him to stop sending MEDs. Or, you can set MED to 0 on your end of the session.

page 181

Internet Routing Architectures, Second Edition

Q—

I am connected to multiple providers. On some occasions, my AS is hit with an enormous amount of traffic that does not belong to my AS. What could be wrong?

A—

You might be advertising routes that you receive from one provider to other providers. Other ASs might be using your AS as transit for their traffic. Make sure that you advertise only your routes to your providers.

Q—

I am multihomed to the same provider. Do I have to worry about advertising the routes I learn from one link back on the second link?

A—

Presumably, BGP policies on your provider's end will detect the routes it is receiving from you that have already passed through its AS and will ignore them. Nevertheless, this is bad practice. By doing so, you're unnecessarily consuming CPU and link resources with useless information; if possible, make sure that you send your own routes only.

Q—

I am a provider, and I have given one of my customers a private AS number. Now, the customer wants to have a different connection with another provider. What would happen if he keeps using the private AS number?

A—

Though this does occur in the Internet today, it is considered a very undesirable configuration. After you advertise this customer's network to the Internet, you are stripping the private AS number and announcing the routes as if they originated from your own AS. If the second provider does the same thing, the customer's networks will be originated from two unique ASs, potentially resulting in routing loops. Also, if any type of aggregation is performed on the customer address before it is announced to other networks, it could result in a morespecific route's being available via one of the two providers, thereby removing the potential for load sharing between the connections. If the customer cannot obtain a global AS number from the appropriate RIR, he should ensure that both providers thoroughly coordinate and provide adequate details regarding load sharing and failover scenarios associated with the configuration.

Q—

I am connected to one provider in San Francisco, and I am advertising my routes via BGP. I am connecting to another provider in LA. Should I get a different AS number?

page 182