| This chapter covers the following key topics about Border Gateway Protocol version 4 (BGP-4): An autonomous system (AS) is a set of devices under common administration. Between two or more autonomous systems, the Border Gateway Protocol advertises network reachability information. The Internet backbone relies solely on BGP to announce and receive IP prefixes, and the only routing protocol that runs between two autonomous systems is BGP. Before BGP, exterior gateway protocol (EGP) was the protocol used between two autonomous systems. EGP was obsoleted by BGP. Why the need for a new protocol? Growing Internet usage in the early 1990s called for a protocol that could provide classless routing and IP prefix advertisement without the concept of network class. Furthermore, this protocol needed to aggregate IP prefixes to shrink the Internet routing table size and robustly advertise a large number of routes to other autonomous systems. BGP offered all that and, among other things, offered mechanisms to control traffic flow in and out of the networks running BGP. In Internet service provider (ISP) networks where revenues are generated by selling Internet access to other small ISPs or to enterprise customers, it is crucial that traffic flows are managed properly. BGP offered ISPs the capability to configure routers with network policies to manage traffic requirements. ISPs make the most use of BGP. Whether it is customer IP traffic destined to the Internet or IP traffic from the Internet to a customer network, BGP allows manipulation of traffic paths to make the best use of the ISP network. Before delving into the various aspects of BGP, you need to (re)familiarize yourself with a few terms: -
IP prefix ‚ This refers to the IP subnet assigned to networks by the official governing body that manages IP addresses. -
BGP feed ‚ This is a commonly used term for a BGP session that provides reachability information of IP prefixes on the Internet. In this context, terms such as full feed and partial feed are also used. Full feed refers to all the Internet prefixes, whereas partial feed refers to a subset of the Internet IP prefixes, based on the traffic requirements. -
BGP peer ‚ BGP peers and BGP neighbors are terms that refer to network devices in the same network that run BGP. -
Router ID (RID) ‚ This is a 32-bit unique identifier representing a BGP speaker. In Cisco IOS Software, the RID is the highest loopback IP address. When loopbacks are not configured, the highest IP address of the interface that is up is taken as the RID. RID can also be manually configured in Cisco IOS. -
Exit point ‚ This is a router that connects two autonomous systems, and traffic comes in and goes out to Internet through the exit point. In most examples, there will be more than one router running EBGP for redundancy and for other requirements. -
Small and large BGP networks ‚ There is no fixed definition of a small or large network. Just one router might exist in the network, or the network might have several hundred routes running the IP routing protocol. -
External BGP (EBGP) ‚ When BGP is run between two autonomous systems, such a BGP session is called External BGP (EBGP). EBGP is primarily used in two different environments: - - Between ISPs and their customers ‚ In this case, customer IP prefixes are advertised through BGP to the ISP and the ISP advertises them to the Internet. However, ISP might advertise full feed or partial feed of the BGP table of the Internet routes to the customer.
- - Between different ISPs ‚ In this case, IP prefixes are advertised to peering ISP connections. This is how all the Internet is glued together.
-
Internal BGP (IBGP) ‚ A BGP session between two routers in the same AS is called an IBGP session. Typically, this is between two or more routers. In IP networks where multiple EBGP peering occurs at multiple exit-point routers with the same or different neighboring AS, it becomes imperative to manage IP traffic coming in and going out to those neighboring autonomous systems. IBGP solves this problem by sharing EBGP feeds between the exit-point routers. IBGP can dictate how traffic will exit the network. For example, an exit-point router can be configured in BGP to send some traffic to the directly connected EBGP link and then send the rest of the traffic to the remote IBGP neighbor. This manages bandwidth requirements of EBGP links and other backbone links. In essence, IBGP plays a significant role in large router-based networks to manage link bandwidth utilization. -
Internet exchange points (IXP) ‚ IXP provides a Real-State in which most, if not all, of the big ISPs exchange BGP routes with each other. -
BGP peering arrangement ‚ In EBGP connections, the two autonomous systems must agree on the kind of BGP peering. The following are the most popular kinds used in the Internet today: - - Transit peering ‚ Suppose that AS A is running EBGP with AS B. If B is configured so that it will pass all Internet traffic from A, B is a transit provider of A. Typically, B will provide a full BGP feed to A.
- - Public peering ‚ An EBGP session at IXP is called public peering.
- - Private peering ‚ An EBGP session on a private link between two autonomous systems is called private peering. It offloads traffic from public-peering locations that are typically congested .
-
Dual or multihoming ‚ When an AS runs more than one EBGP session with the same or different AS, it is considered dual or mutlihomed to that AS. Dual- homed networks might have single or multiple routers in the AS. This provides redundant connections to the Internet and also provides load sharing. -
BGP policies ‚ These are BGP rules designed to predict how BGP influences traffic-flow policies coming in or going out of the network. Policies are either configured or are taken from the default behavior of BGP protocol. -
Administrative distance (AD) ‚ Cisco IOS Software assigns an AD to each protocol. AD has local significance in the router and is not exchanged with any other routers. In Cisco IOS Software, EBGP and IBGP have an AD of 20 and 200, respectively. When a prefix is learned by two different protocols in the same router, AD does the tie breaking and the lower AD prefix is installed in the IP routing table. Cisco IOS Software also enables you to reconfigure AD values under the routing protocol command set using the distance command. -
BGP best path ‚ By definition of RFC 1771, BGP must decide on a single best route out of many to install in the routing table. If BGP receives multiple advertisements from multiple neighbors for the same prefix, it must decide on a single best route through BGP best-path selection, discussed later in this chapter. It is this best route that BGP installs in the IP routing table and advertises to other BGP neighbors. -
Hot potato ‚ A commonly used term for a BGP policy that governs that traffic will exit the AS from the closest exit-point router. -
Cold potato ‚ A commonly used term for a BGP policy that governs that traffic will be delivered through the path that is closest to the destination. Optimal routing can be viewed as cold potato routing. Figure 14-1 shows that AS A, C, and D are running EBGP sessions with AS B. Routers AS B ‚ namely, R1, R2, R3, R4, and R5 ‚ are shown to run IBGP with each other, and they are fully meshed with each other. AS A is dual homed to AS B for redundancy and load sharing. AS A has one high-bandwidth link and one low-bandwidth link to AS B. In addition, AS B is providing transit services to AS C, and AS C also has a private peering session with AS D. Figure 14-1. Sample BGP Network  Figure 14-1 provides a simple view of an ISP B network. All such ISPs connect with each other to form this Internet. These ISPs might connect at IXP, or they might have private peering with each other, like AS C and AS D do in this figure. Figure 14-1 shows that all autonomous systems except for AS C must go through AS B to reach other networks. AS C may use its private peering link with AS D for all Internet traffic or some other traffic, depending on the kind of BGP feed (full or partial) exchanged. The kind of BGP feed from AS D to AS C and local BGP policy of C dictates how traffic goes out of the C network. This is one example of BGP policy. In another example from Figure 14-1, AS A is dual homed with AS B but has one high-bandwidth link and another low-bandwidth link. AS A might use a high-bandwidth link to its full capacity and might not use low bandwidth at all; AS A can choose to use a low-bandwidth link for some traffic, and the rest of the traffic can go on the bigger link. All these policies and requirements can be serviced by BGP, and that makes usage of BGP so important and powerful. |
