Solve the horizontal segmentation of iBGP and BGP routing principles

solve IBGP Horizontal segmentation problem of

IBGP Horizontal segmentation ： To prevent the generation of loops

Method 1： Routing reflectors (RR)

The route reflector can reflect the learned IBGP Routing information .

Specify a router called Routing reflectors （RR） When the , Must be in his IBGP Selecting one or more devices in the peer relationship is called his Customer （client）,RR The system formed with customers is called Reflection cluster （Cluster）. Each reflection cluster will use RR Of RID As a cluster ID. The rest is not called RR Customer's IBGP Peer relationship , We will call it Non customers .

Reflection rules of routing reflectors ：

1, When a route reflector learns a route from its client , Then he will reflect this routing information to all his customers and non customers

2, If the route reflector learns a route from its non customers , Then he will reflect this routing information to all his customers, but not to non customers .

3, When the router performs route reflection , He only uses the best BGP Route for reflection .

The route reflector reflects the route and then breaks IBGP The horizontal division of , It may cause the appearance of the loop . The routing reflector is to prevent the appearance of the loop , Two properties are introduced Originator_ID（ Originator ID）,Cluster_list（ Cluster list ）

Originator_ID（ Originator ID） When a routing message comes to the routing reflector , The reflector needs to reflect this route information , This attribute will be added to the routing information before reflection , This The value of the attribute is that of the notifier who received this route information RID, after , Other routing reflectors receive a routing message if it contains O_ID, Then he will not change this attribute . If a device receives a routing entry, the originator ID For their own local RID, You will no longer learn this routing information , Play the effect of preventing loops .

Cluster_list（ Cluster list ） If in a AS among , There are multiple reflections , Then there must be multiple reflection clusters , Then each RR When reflecting the routing information, local clusters will be added to the cluster list attribute ID. When a device receives a reflected routing message , The cluster list attribute contains local clusters ID, You will no longer learn this route , To prevent the generation of loops .

[r3-bgp]peer 2.2.2.2 reflect-client --- Appoint R2 As a reflective customer , At the same time R3 Automatically become RR

 

  Method 2： The federal

The federal ： Will be the same AS In the middle of IBGP The peer relationship is changed to EBGP Peer relationship . This requires different IBGP Peers are divided into different members AS among , member AS Between the establishment of EBGP Peer relationship is a special EBGP Peer relationship , They only follow EBGP Transitivity of peers , From the perspective of the content of transmission , Still need to follow AS-BY-AS The rules , Because from a macro point of view , They are still in the same AS among .

The federal approach is also equivalent to breaking IBGP The horizontal division of , Then there may also be loop questions topic , You need to consider anti ring . The anti ring method can be directly used for reference EBGP Horizontal segmentation , stay AS_PATH Add members to the attribute AS No. to prevent backtransmission . Be careful , To distinguish members AS And normal AS, member AS The number will be enclosed in parentheses .

  Configuration of federation ：

[r2]bgp 64512--- Federate device startup BGP The process needs to be started with a small number

[r2-bgp]confederation id 2--- Declare the large size of federal members

[r2-bgp]peer 3.3.3.3 as-number 64512--- Between members of the Federation IBGP Peer relationships need to be established with a trumpet

[r3-bgp]confederation peer-as 64513--- In need of federation EBGP A peer-to-peer device needs to declare a neighbor member AS Number

[r3-bgp]peer 4.4.4.4 as-number 64513--- Federal EBGP Peer relationships use trumpets to build neighbors

[r3-bgp]peer 4.4.4.4 ebgp-max-hop --- Because of the federal EBGP Peer relationships need to follow EBGP Transitivity of peers , therefore , Its TTL The value is defaults to 1, When building neighbors through indirect connection, it is necessary to TTL Only when the value is modified can the neighbor be built normally

BGP The principle of road selection

BGP The prerequisite for choosing a route --- Discard all unavailable routing information

1. optimization PV Routes with large attribute values

hypothesis R1 Publish two routing entries on 1.0.0.0 10.0.0.0 ,R4 adopt BGP Peer relationship , It can be downloaded from R2 and R3 I learned two ways to go 1.0.0.0 10.0.0.0 The routing entry for . When adding a table to a route, only the best one will be selected , Therefore, the routing can be affected by configuration .

Method 1 ： Global modification  

[r4-bgp]peer 3.3.3.3 preferred-value 100

This 100, It's modified PV value . The meaning of this order is ： As long as it is 3.3.3.3 Inform to R4 The routing , These routes PV The fields are all changed to 100.

Method 2 : Fine control

1. Grab traffic

[r4]ip ip-prefix PV permit 10.0.0.0 24

2. Routing strategy

[r4]route-policy PV permit node 10

[r4-route-policy]if-match ip-prefix pv

[r4-route-policy]apply preferred-value 100

[r4-route-policy]q

[r4]route-policy PV permit node 20--- Be careful ： Empty meter must be configured to discharge all remaining traffic

3.BGP Call policy

[r4-bgp]peer 3.3.3.3 route-policy PV import   This piece can only be made into The direction of the , because PV Attribute cannot be transferred Deliver

2. optimization LP Routes with large attribute values

LP --- Local priority ( Local ： At present AS In the transfer )--- The default value is 100, The bigger the better . This property can be used in IBGP Pass between peers . Routing information is being passed to EBGP Peer will not carry this attribute .

Method 1 ： Global modification

[r3-bgp]default local-preference 200---R3 Then send it to IBGP Peer routing LP Properties will be changed to 200

Method 2 ： Fine control

3. According to the priority of the route entry publishing method

The priority of self initiated routes is higher than that learned from peers , The priority relationship of the route originated by itself is manual aggregation > Automatic aggregation > network > import

4. optimization AS_PATH Attribute the shortest route

AS_PATH Attribute when selecting a route , Be careful ：1. When performing route aggregation , If activated AS_SET After keyword , Will be different AS Aggregate detailed routes , Will be in AS_PATH Attribute carries all details AS Of AS Number , And enclose it with braces . When comparing routes , The contents of braces are treated as a AS No . 2, In the union , We also need to make AS_path To protect the environment , We will use parentheses to put AS Put the numbers around , When comparing routes , Don't count the contents in parentheses .

5. Choose the route with the best origin code attribute . Priority of origin code ---I > E > ?

OGN Origin code ： Used to identify the origin of routing entries

I --- Represents that the routing information originates from IGP agreement （ Not limited to IGP agreement , Including static , Direct connection ）, Represents that the route entry originated from AS Inside --- adopt network The origin code of the published routing information is I

E --- Represents that the routing information originates from EGP agreement ---EGP refer to BGP The external gateway protocol used before

？--- Routes learned through other than the above two methods --- The route origin codes of republishing and importing are ？

6. optimization MED Routes with small attribute values

MED --- Multi exit authentication attribute ---MED Value is the cost value that inherits the routing entry in the routing table , It can reflect the arrival of the target network segment in this AS Internal overhead size . Through this parameter , other AS The flow is entering this AS You will be able to choose the less expensive entrance to enter . By modifying this attribute value , We can also influence others AS Flow into this AS The path of . If it passes locally IBGP The peer learns a BGP route , There is MED value , Then you need to pass this route to the local EBGP peers , Then he will no longer carry this MED value . This setting may cause poor route selection . therefore , Suggest , If there are multiple groups of boundary devices, you need to send routes to EBGP Peer relationship , All routing information needs to be published .（ Each set

All equipment needs to be released ）

7,EBGP Peer routing is better than IBGP peers

8. Preferred to NEXT-HOP Of IGP The measure is small (cost) The routing ：

9. choose CLUSTER_LIST The shortest routing information

10. optimization RID(O_ID) Minimal device advertised routing

First look at O_ID, No originator looks RID

11. It is preferred to have a minimum IP Route of peer advertisement of address