BGP Multihoming for Enterprise ISPs

1. The "What" and "Why" of BGP Multihoming

Multihoming is the practice of connecting your network to more than one ISP. This design provides two primary benefits:

• Redundancy: If one ISP connection fails, traffic can automatically fail over to the other, ensuring continuous internet connectivity.
• Traffic Engineering: You can intelligently route traffic over the best-performing path or load-balance across links to optimize cost and performance.

The typical architecture involves establishing eBGP peerings with each ISP, while running iBGP between your own edge routers to share routing information and enforce consistent policies.

The Critical Risk: Becoming a Transit AS

When you connect to two ISPs, you create a potential path between them through your network. If you're not careful, your routers could start advertising routes learned from ISP1 to ISP2. This could cause internet traffic from ISP2 destined for networks at ISP1 to flow through your enterprise network. This is known as becoming a transit autonomous system. This is almost always undesirable, as it can saturate your internet links and overload your routers with traffic that has nothing to do with your business.

To understand this risk and the fundamental solution, please read the introductory section of the following Cisco guide.

As the document explains, the key is to filter your outbound BGP advertisements so that you only announce your own, locally originated prefixes. The regular expression ^$ is a powerful tool for this. It matches an empty AS-Path, which is characteristic of routes originating within your own AS. By applying a route-map with this filter to your eBGP neighbors, you ensure you never re-advertise routes learned from one ISP to another.

Multihoming Strategies

An enterprise doesn't have to accept the entire internet routing table (over 900,000 routes!) from its ISPs. There are several common strategies:

• Full Routes: Accept all BGP routes from all ISPs. This provides optimal routing decisions but requires powerful routers with significant memory.
• Partial Routes: Accept only routes from your ISPs and their direct customers. This reduces the routing table size while still providing good path information for nearby networks. A default route is still needed for the rest of the internet.
• Default Routes Only: The simplest approach. You only accept a default route from each ISP. This is easy to manage but gives you no ability to make granular path selections.

Your choice of strategy depends on your technical requirements, hardware capabilities, and business needs.

2. Advertising Your Network Prefixes

For multihoming to work, you must advertise your own public IP address block to your ISPs. This is done using the BGP network command. However, this command works differently in BGP than in IGPs like OSPF or EIGRP.

Let's watch a video that explains this crucial concept and demonstrates the configuration.

As the video demonstrated, for network 198.51.100.0 mask 255.255.255.0 to work, the route 198.51.100.0/24 must exist in the router's IP routing table. A common and clever technique to ensure this, especially for an aggregate prefix, is to create a static route pointing to the Null0 interface (e.g., ip route 198.51.100.0 255.255.255.0 Null0). This makes the route available in the routing table for BGP to advertise, without affecting traffic forwarding for more specific subnets.

3. Controlling Traffic Flow

Once your peerings are up and you're advertising your prefixes, the next step is to control how traffic flows. This is done differently for outbound and inbound traffic.

Outbound Traffic: Your Choice with Local Preference

You have complete control over how traffic leaves your AS. The primary tool for this is the Local Preference attribute.

• It is a BGP attribute with a default value of 100.
• It is only exchanged between iBGP peers within your AS. It is not sent to eBGP peers (ISPs).
• A path with a higher local preference is preferred.

By setting a higher local preference for routes learned from one ISP, you can instruct all routers within your AS to prefer that ISP as the exit point for those destinations. This is typically done with a route map applied to incoming advertisements from an eBGP peer.

Inbound Traffic: Their Choice, Your Influence with AS-Path Prepending

You cannot directly control how traffic enters your AS. That is decided by other networks on the internet. However, you can influence their decision. One of the most common ways to do this is by manipulating the AS_PATH attribute.

A shorter AS_PATH is one of the most important factors in BGP path selection. By making the path through one of your ISP connections appear longer, you can make it less attractive to other networks. This technique is called AS-Path Prepending. You artificially lengthen the AS-PATH by adding your own AS number to it multiple times.

The following video explains this concept and shows a real-world configuration.

A key takeaway is that you use a route-map applied to your outbound advertisements to an ISP. For the path you want to make less preferred, you prepend your AS number. For example: set as-path prepend 65001 65001 65001.

4. Case Study: A Complete Multihoming Configuration

We've covered several individual concepts. Now, let's tie them together by examining a complete, practical design example from Cisco's documentation. This case study demonstrates how an enterprise (AS100) connects to two ISPs (AS200 and AS300), uses OSPF as its internal routing protocol, and applies BGP policies to manage traffic.

This example effectively demonstrates a common multihoming strategy:

• Use ISP-A (via RTA) as the primary path for most internet traffic (local preference 200).
• Use ISP-B (via RTB) as the preferred path for specific routes local to that ISP (local preference 300, which is higher).
• Inject default routes into the internal network from both edge routers, with different metrics, to provide a clear primary and backup path.

This is a sophisticated design that combines redundancy with policy-based performance optimization, showcasing the power and flexibility of BGP.

Conclusion

In this lesson, we moved from the internal workings of BGP to its most common real-world application: enterprise multihoming. You now have the foundational knowledge to design and implement a resilient and intelligent internet edge for an enterprise network.

Key Takeaways:

• Multihoming provides internet redundancy and allows for traffic engineering.
• It is critical to filter outbound advertisements (e.g., using ip as-path access-list permit ^$) to prevent your network from becoming a transit AS.
• Local Preference is the primary tool to control outbound traffic flow, steering traffic to the preferred ISP exit point.
• AS-Path Prepending is a key technique to influence inbound traffic flow by making a path appear less desirable to external networks.
• A combination of these policies allows you to create sophisticated traffic engineering solutions tailored to your business needs.

Preview of the Next Lesson:

Now that you have configured a complex, multihomed BGP environment, the next logical step is to learn how to fix it when it doesn't work as expected. In our next lesson, we will cover the learning outcome: Troubleshoot BGP peering issues and route advertisement problems. We will dive into show and debug commands to diagnose and resolve common BGP failures.