BGP details-Border Gateway Protocol (4)

10. Conclusion
The BGP protocol provides high control and flexibility for Inter-Domain Routing, implements policy constraints, and avoids routing loops. The guidelines described here provide a starting point for BGP use. As BGP grows, it will provide more advanced and manageable routing methods.
　　
Appendix A. Relationship between BGP and IGP
　　
This section describes how to exchange routing information between BGP and IGP. These methods are not proposed as part of standard BGP. These methods are only information supplements. When an application introduces IGP information, you may need to consider these methods.
This is an overview for general IGP.
The relationship between BGP and a specific IGP is not discussed in this section. The specific IGP method should be discussed in other documents and should be standardized in the future.
　　
Overview
　　
According to the definition, all transition as must be able to transport the traffic that comes from or finally the. This requires a certain relationship and coordination between BGP and the Internal Gateway Protocol (IGP) used by the specific. In general, traffic originating from external as will be routed through the Internal Gateway (only IGP is supported) and the Border Gateway (both IGP and BGP are supported) at the same time ). All internal gateways receive information about external routes from more than one border gateway through IGP.
Based on the mechanism of transmitting BGP information within an as, the consistency between BGP and IGP should be paid special attention, because the status change may spread in the as at different speeds. There may be time windows. The two time periods are: Obtain New BGP Route information from another border gateway (B) In the same as in a border gateway (; as IGP can route traffic to the Border Gateway (B ). The error routing and "black hole" can occur between these two moments.
To minimize this routing problem, before all the internal gateways of AS are ready to send traffic destined for the external IP address to the correct egress Border Gateway (B), Border Gateway () routes destined for the external target network through the border gateway (B) should not be broadcast to the external peer. In other words, an internal route must converge to an appropriate egress gateway before it can broadcast a route through the egress gateway to an external peer.
　　
A.2 method for achieving stable relationships
　　
The following discusses several technologies that can realize the stable relationship between BGP and IGP in.
　　
A.2.1 pass BGP information through IGP
　　
BGP provides its own mechanism to transmit BGP information within the AS. Similarly, As long as IGP supports flooding of all routing information (it provides a mechanism to distribute BGP Information) and a convergence of a channel (which makes the mechanism efficient atomic), it can be used to transmit BGP information. If an IGP is used to transmit BGP information, the aforementioned asynchronous time will no longer exist, because as's internal BGP information transmission is synchronized with IGP, IGP will converge more or less with the arrival of new route information. Note that IGP only ships BGP information and does not parse or process it.
　　
A.2.2 Internal Gateway Protocol marked
　　
When outbound routes are transmitted within the AS, some IGP can mark them with their egress points. Each Border Gateway should use the same tag to declare external routing information (received through BGP ), whether the route information enters IGP or is passed to another internal peer (with the same as number ). The tags generated by the border gateway must be unique to the specific border gateway. Different border gateways must use different tags.
All border gateways in an as must comply with the following two rules:
1. If boundary gateway A receives information from an internal peer claiming that some target networks are inaccessible, it must pass this information to all external peers.
2. if Border Gateway A receives information about some reachable destination network X from an internal peer, it must have an IGP route to X, in addition, the routes of IGP and BGP are marked accordingly. Otherwise, the information cannot be transmitted to any external peer.
These rules ensure that route information is not declared unless the IGP supports it correctly. It can also partially avoid "black holes ".
One way to mark BGP and IGP in as is to use the IP address of the egress Border Gateway. In this case, the "Gateway" domain of the bgp update message is used as a tag.
　　
A.2.3 Encapsulation
　　
Encapsulation provides the simplest mechanism (about the relationship between IGP and BGP) for transit as traffic ). In this way, the traffic is encapsulated in an IP packet whose address is the egress gateway. The only requirement for IGP in this method is that it supports routes between the same as border gateways.
Go to the address of exit gateway A of some external target network X and declare it through the BGP identifier in the BGP open message sent by a to other border gateways in the same. In order to route traffic to the target X, each Border Gateway in as encapsulates the IP data packets sent from the traffic to the IP address of gateway. Then, Gateway A unpacks and sends the original data packet to the corresponding gateway of the external.
Because encapsulation does not rely on IGP to transport external routing information, there is no need to synchronize between BGP and IGP.
Some methods are used to identify packets containing encapsulated IP addresses. For example, the IP protocol type code must be defined before use.
Note: If the length of an encapsulated packet is very close to that of MTU, the packet will be split by the gateway that executes the encapsulation.
　　
A.2.4 distributed BGP
　　
If all routers in an as are BGP spokespersons, there is no need to contact BGP and IGP. In this case, all routers in the as have all the information about the BGP Route. IGP is only used for routing in as, and BGP routes are no longer imported into IGP.
If the routers work in this way, they must be able to perform a loop query on the route table. The first query uses a BGP Route to establish the egress router. The second query determines the IGP path to the egress router.
In this case, IGP does not carry any external information. Once a BGP spokesman obtains new information about the route, the routers in the as will soon gather. Because there is no IGP convergence delay, an application can broadcast these routes without delay.
　　
A.2.5 other cases
　　
Some as IGP may neither transmit BGP information nor mark external routes (such as rip ). In addition, encapsulation may be unfeasible or unacceptable. In these cases, the following two rules must be observed:
(1) If Border Gateway A receives information from an internal peer that claims that some target networks are inaccessible, it must pass this information to all external peers.
(2 ). if Border Gateway A receives information about some reachable target network X from an internal peer, it must have an IGP route to X and have enough time to converge the IGP route, otherwise, the information cannot be transmitted to any external peer.
The preceding rules only indicate the necessary (but not sufficient) Conditions for passing BGP Route information to other. Compared with the labeled IGP, these rules cannot guarantee that the internal route to the egress gateway exists before the route is routed to another.
If the IGP convergence time is smaller than a small value of X and the time window caused by the non-synchronization of IGP and BGP is smaller than X, the route instability is abnormal for a short time, which can be ignored. The reasonable value of X is not yet studied, but should be less than 1 second.
If the IGP convergence time cannot be ignored, different methods are required. The mechanisms and technologies suitable for this situation need to be studied in depth.