Last Updated on Wed, 05 Jul 2023 |Network Design
The generic meaning of multihoming is to "provide more than one connection for a system to access and offer network services." The term multihoming is used in many specific ways also. A server, for example, is said to be multihomed if it has more than one network layer address. Content delivery networks can multihome application layer data and services.
The term multihoming is increasingly being used to refer to the practice of providing an enterprise network more than one entry into the Internet. Redundant entries into the Internet provide fault tolerance for applications that require Internet access. An enterprise network can be multihomed to the Internet in many different ways, depending on a customer's goals. Figure 5-14 and Table 5-2 describe some methods for multihoming the Internet connection.
Figure 5-14. Options for Multihoming the Internet Connection
Table 5-2. Description of Options for Multihoming the Internet
Connection
Table 5-2. Description of Options for Multihoming the Internet
Connection
Enterprise | Number of Connections to the Internet | Number of ISPs | Advantages | Disadvantages | |
Option A | 1 | 2 | 1 | WAN backup; low cost; working with one ISP can be easier than working with multiple ISPs. | No ISP redundancy; router is a single point of failure; this solution assumes the ISP has two access points near the enterprise. |
Option | 1 | 2 | 2 | WAN backup; | Router is a single |
B | low cost; ISP redundancy. | point of failure; it can be difficult to deal with policies and procedures of two different ISPs. | |||
Option | 2 | 2 | 1 | WAN backup; | No ISP |
C | especially good for geographically dispersed company; medium cost; working with one ISP can be easier than working with multiple ISPs. | redundancy. | |||
Option | 2 | 2 | 2 | WAN backup; | High cost; it can |
D | especially good for geographically dispersed company; ISP redundancy. | be difficult to deal with policies and procedures of two different ISPs. |
In the case of Options C and D, the goal might be to improve network performance by allowing European enterprise sites to access the Internet using the Paris router and North American sites to use the New York router. This can be accomplished by correctly configuring a default gateway on end stations and a default route on enterprise routers in Europe and North America. (A default route specifies where a packet should go if there is no explicit entry for the destination network in a router's routing table. Default route is also sometimes called the gateway of last resort.)
Your customer might have more complex goals than the simple goal in the previous paragraph. Perhaps your customer wants to guarantee that European enterprise sites access North American Internet sites via the New York router. A parallel goal is that North American enterprise sites access European Internet sites via the Paris router. This could be a reasonable goal when a constant, low latency is required for an application. The latency is more predictable if the first part of the path is across the enterprise intranet instead of the Internet. This goal is harder to meet than the first goal, however. It requires that the enterprise routers understand routes from the ISP and set preferences on those routes.
A related goal is to use the "best route" across the Internet to the sites that the enterprise users most rely on. Unless an enterprise contracts (and pays) for end-to-end managed quality of service (QoS), this goal cannot be met. The routing protocol used on the Internet, BGP, doesn't offer route optimality. Its only purpose is to provide reachability and stability in the global routing system. Intermediate providers with whom an enterprise has no business relationship don't care if the enterprise's traffic follows optimal routes, nor do they have any incentive to do so.
Another more complex goal is to guarantee that incoming traffic from the Internet destined for European enterprise sites uses the Paris router and incoming traffic for North American enterprise sites uses the New York router. This goal requires the enterprise routers to advertise to the Internet routes to enterprise sites. The routes must include metrics so that routers on the Internet know the preferred path to sites on the enterprise intranet.
One other caveat when an enterprise network is multihomed is the potential to become a transit network that provides interconnections for other networks. Looking at the pictures in Figure 5-14, consider that the enterprise router learns routes from the ISP. If the enterprise router advertises these learned routes, then it risks allowing the enterprise network to become a transit network and being loaded by unintended external traffic. When an enterprise network becomes a transit network, routers on the Internet learn that they can reach other routers on the Internet via the enterprise network. To avoid this situation, enterprise routers should advertise only their own routes. (Alternatively they cannot run a routing protocol and depend on default and static routing.)
In general, multihoming the Internet connection can be challenging if a customer's goals are complex. Encourage your customers to simplify their goals to ensure ease of implementation, scalability, availability, and affordability. If the main goal is high availability, don't assume that this means more redundancy is required. According to Howard Berkowitz in his book WAN Survival Guide, "Uncontrolled increases in redundancy lead to uncontrolled increases in complexity, and may actually decrease availability." See Berkowitz's book and other books listed in Appendix B, "References and Recommended Reading," for more information on maximizing the availability of Internet access.
Continue reading here: The Service Provider Edge
Was this article helpful?