There are several factors that should be taken into consideration when connecting the enterprise network to an Internet Service Provider (ISP). These factors include the following:
- Whether or not you want to use dynamic or static routing
- Whether or not you have a public AS number
- Whether or not you have your own registered public IP address space
- The acceptable or desired level of Internet redundancy
- What type(s) of link(s) you will need to connect to the ISP
It is important to understand that you do not need to enable BGP in order to connect to an ISP. As will be explained later in this section, static routing can be used to connect to an ISP. If you do want to use BGP for greater policy control, your implementations options will vary depending on whether or not you have a public autonomous system number assigned to your organization. Again, it is not mandatory that you have a publically assigned AS to connect to the ISP; however, having a publically assigned AS provides the most flexibility.
Consideration should be given to the acceptable level of Internet redundancy. For small enterprises, Internet connectivity may not be critical; however, for large enterprises that host their own Internet-facing servers, for example, both reliable and redundant Internet connectivity would be critical to the organization. The enterprise BGP implementation should therefore take this into consideration when connecting to Internet Service Providers.
There are three types of implementations that can be used when connecting to Internet Service Providers. These three implementation or deployment models are:
- Single-homed implementations
- Stub Multi-homed implementations
- Multi-homed implementations
A single-homed implementation is one where the enterprise network has a single connection to the Internet Service Provider. A multi-homed implementation is one where the enterprise network has more than one path to one or more Internet Service Providers. While single-homed implementations are the easiest method of connecting to an ISP, multi-homed implementations provide greater reliability and allow for optimal routing.
Multi-homing provides a greater level of redundancy by eliminating single points of failure. This allows for much a more reliable service than that afforded by single-homed implementations. In addition, multi-homing provides network engineers with greater routing policy control, allowing for optimal routing into and out of the enterprise network. These deployment models are described in the following sections.
Single-Homed (Stub) Implementations
A network is single-homed when it has a single connection to the ISP. This implementation is also commonly referred to as a stub network. Figure 6-2 shows a single-homed or stub enterprise network that is connected to an Internet Service Provider:
Fig. 6-2. Single-Homed (Stub) Network Implementation
Referencing Figure 6-2, the enterprise network is single-homed to an Internet Service Provider which may be directly or indirectly connected to the Internet. Stub implementations are commonly used by small businesses f
or which Internet access is not critical as they provide a low-cost solution for connecting the enterprise network to the Internet. The ISP advertises the stub network subnet to the rest of the Internet community.
In addition to being a low-cost solution, single-homed or stub implementations are also simple to deploy as it is not necessary to implement Border Gateway Protocol on the customer edge router. Instead, a single static default route to the ISP router for all Internet-bound traffic can be used. The provider then uses BGP to advertise the network to the Internet, and all policy control is solely in the hands of the provider. In other words, the enterprise cannot influence how the stub network is advertised to the rest of the Internet community.
It is important to understand that while single-homed implementations offer little redundancy due to both the edge router and the circuit being single points of failure, an Internet outage due to a circuit failure can be mitigated against by provisioning multiple circuits between the enterprise (customer) edge router and the provider edge router, as illustrated in Figure 6-3:
Fig. 6-3. Circuit Redundancy for a Single-Homed Connection to a Service Provider
Referencing Figure 6-3, multiple (two) circuits between the customer edge and provider edge routers reduce the likelihood of an outright Internet outage due to a single circuit failing. In implementations such as the one shown in Figure 6-3, multiple default static routes can be configured on the customer edge router or the physical circuits can be bonded into a single logical link, using WAN technologies such as PPP Multilink for example.
Stub Multi-Homed Implementations
A stub multi-homed network is a network that is connected to two or more edge routers that belong to the same Internet Service Provider. Stub multi-homed networks can have a single edge router or multiple edge routers for redundancy. Most stub multi-homed implementations have the following characteristics:
- BGP is used to exchange routing information with the ISP
- A private BGP AS number, i.e. 64,512 to 65,535 is used
- The upstream ISP advertises the stub network to the Internet
- BGP is used for greater routing policy control
Figure 6-4 illustrates a stub multi-homed network using a single edge router connected to two provider edge routers belonging to the same Internet Service Provider:
Fig. 6-4. Single Edge Stub Multi-Homed Networks
Referencing Figure 6-4, BGP is used to the exchange routing information between the enterprise and the ISP. It should be noted that static routes can also be used between the enterprise edge router and the ISP edge router; however, static routes do not provide the flexibility provided by BGP and do not afford the enterprise the same granular routing and traffic flow controls.
Because the enterprise is connected to a single ISP, a private AS number can be used, negating the need for the enterprise to obtain a public AS number from a regional registry. As was described earlier in this chapter, the ISP simply removes the private AS number from routing updates to upstream autonomous systems or the Internet.
In order to eliminate the single enterprise edge router as a single point of failure, the stub multi-homed implementations can also be deployed using multiple edge routers. Multi edge stub multi-homed networks are implemented by connecting more than one edge router to multiple provider edge routers (belonging to the same ISP) as illustrated in Figure 6-5:
Fig. 6-5. Multi-Edge Stub Multi-Homed Networks
Figure 6-5 shows the same enterprise network depicted in Figure 6-4; however, this time, there are two edge routers being used instead. While the same basic configurations would be used when using two customer edge routers instead of one, if there are any Layer devices that will be providing transit service between the two edge routers, the enterprise should implement an internal BGP (iBGP) mesh between these devices and the edge routers. This ensures that intermediate devices have routing information for the BGP networks being advertised between the edge routers, while negating the need to redistribute BGP routes into the IGP. It also prevents the black-holing of network traffic. Such an implementation is shown in Figure 6-6:
Fig. 6-6. Using iBGP in Multi-Edge Stub Multi-Homed Networks
Referencing the network topology illustrated in Figure 6-6, a full iBGP mesh is implemented the enterprise (AS 650001) allowing the intermediate router providing transit services between the two edge routers to learn about the routes received by the edge routers from the ISP via BGP, versus through the redistribution of these routes into an IGP on either edge router. This is most commonly implemented when the ISP is advertising a partial or full Internet table, which can be well over a hundred thousand prefixes, to the enterprise. This large number of prefixes can quickly overwhelm an IGP. However, if the ISP is advertising only a single default route, this can be redistributed into the IGP and propagated downstream through the enterprise network.
NOTE: The options that are available when receiving routes from the Internet Service Provider are described in detail in the following section.
Multi-homed implementations
Standard multi-homed implementations are commonly used by medium-to-large enterprise organizations. A standard multi-homed network is a network that is connected to multiple upstream Internet Service Providers. Most standard multi-homed implementations have the following characteristics:
- BGP is used to exchange routing information with the ISP
- A public BGP AS number, i.e. 1 to 64,511, is used
- Public IP address space is used
- BGP is used for greater routing policy control
As with stub multi-homed implementations, standard multi-homed implementations also use external BGP to exchange NLRI between enterprise and provider autonomous systems. However, unlike stub multi-homed implementations, standard multi-homed implementations use a publically registered autonomous system number, as well as IP address space. Because they use publically registered autonomous system numbers and global addressing, standard multi-homed implementations can be used to peer to more than one Internet Service Provider. Figure 6-7 shows a standard multi-homed implementation using a single edge router:
Fig. 6-7. Standard Single-Edge Multi-Homed Network
Referencing Figure 6-7, the enterprise network is multi-homed to two different ISPs using a single edge router. This solution provides the enterprise with redundancy against a single circuit or provider edge router failure. This implementation also allows BGP to be used for greater path control for packets originated from the enterprise network to the Internet, as well as for traffic from the Internet to the enterprise.
While this implementation does provide circuit and provider edge redundancy, the single edge enterprise router still represents a single point of failure. A more commonly implemented solution, which shows multiple enterprise edge routers, is shown in the diagram in Figure 6-8:
Fig. 6-8. Standard Multi-Edge Multi-Homed Network
Figure 6-8 shows a standard dual-homed network connected to two different ISPs. Within the enterprise network, iBGP may be used to ensure that any transit devices between the two enterprise edge routers known about the BGP routes received by the edge routers as was also stated in the section on stub multi-homed implementations.






