It seems that you are describing two different things, the forming of BGP peerings with neighbours, and the choice of the next-hop BGP router.
Concerning choosing neighbours, in order for BGP to function correctly, you don’t really have a choice of what neighbour peerings to make. iBGP needs to have a “full mesh” of iBGP neighbours within the AS, which means that each and every iBGP router in an AS must be neighbours with every other iBGP router. As for eBGP neighbours, you have more freedom to choose which routers will be neighbours with witch. But the choice of routers as neighbours affects the network topology and does not directly affect the route a particular prefix will take. That is defined by the next-hop IP.
You can indeed affect the choice of the next-hop IP, which will affect the path a packet will take using BGP, and this is achieved using the BGP attributes. There are various ways to do this, and these are described in the BGP attributes and path selection lesson:
Take a look at this lesson and if you have any more questions, please let us know!
Unless I missed it, the question as to why the route with 203.202.143.33 was chosen was not discussed - and the result shown is based upon the following…
I’m assuming that the Cisco routers in this example use the (Cisco, non-IETF) default behavior where the non-presence of any MED value is deemed “better” than the presence of a specific Med value.
If true, in this instance the route selection would come down to a choice between 203.202.143.33 and 203.202.143.34 - since these are the only candidate routes than have no MED values. I then assume that the subsequent path selection rules are a tie until we get to either the neighbor router ID or (if that’s a tie as well) the neighbor IP address for 203.202.143.33 would be preferred over the alternative address of 203.202.143.34.
Is that correct?
BTW- great site! Great course! Great article! Many thanks.
According to Cisco, paths that have no MED are actually assigned a MED of 0 and are evaluated using that value, making a MED of 0 preferrable to any other MED value since it is less than any other value a path can have. You can always change this behaviour by using the bgp bestpath med missing-as-worst command to employ the opposite behaviour. You can find more on this at this Cisco documentation. So any non-existent MED in the output you see is equivalent to a MED of 0.
With the information that is given, yes, your description is correct.
It has not yet been added, but Rene will get back to you on this. In the meantime, take a look at this Cisco documentation that describes the Accumulated IGP feature for BGP:
How BGP selects path if I advertised routes from same AS numbers where BGP detects two similar paths with nearly the same local preference, and weight as well?
BGP will always follow the path selection process based on a series of attributes that it compares. These attributes are compared in a specific order and are used as tie-breakers. If the attribute is the same for two or more paths, then the next attribute is compared. If all else fails, the very last tiebreaker is the neighbor IP address, where the path with the lowest neighbor IP address is preferred. This will always be a tie-breaker because the IP addresses of neighbors cannot be the same.
The first attribute that is examined is weight. If it is the same, then the local preference will be examined. If the local preference is “nearly the same” then it is not the same, so the highest local preference will be chosen.
Take a look at this lesson to see all of the attributes and the order in which they are examined.
How do you manipulate the inbound and outbound flow of traffic in BGP?
If we use weight to manipulate the outbound flow of traffic then why are we setting this as ‘in’ direction for the neighbor statement and similarly for the inbound flow of traffic we set it in ‘out’ direction.
Can AS-Path be used to manipulate both inbound and outbound?
Take a look at this post, I believe it will clarify the issues with “in” and “out” that you mention here:
But how else do we manipulate incoming and outgoing traffic? You have ultimate control over your outbound traffic flow using BGP. However, you do not have ultimate control over your incoming traffic. You can influence it in various ways, but you can’t ultimately control it. More info about this can be found at the following post:
If you have any further questions, feel free to ask!
In all cases, if you receive a particular prefix from multiple sources, then you still go through the BGP attributes in order, so that you can determine which path is best, and thus, which path to put in the BGP table.
By default, BGP will select only a single path for a prefix. You can see this from the very last attribute that is examined. If all other attributes are the same, at the very end, a BGP router will prefer the path with the lowest neighbor IP address. This is the last tiebreaker, and the neighbor IP addresses will never be the same, so one will always win out over the other.
If you want BGP to perform load balancing, you will have to configure it to do so. This can be done using a feature called Multipath load sharing, and you can learn about it here:
If you have routing information to a particular destination that is coming from several routing sources such as different routing protocols, then yes, the first thing that would be examined is the administrative distance.
Secondly, for all routing information you are receiving about a particular destination, the prefix length must be the same, otherwise, you are looking at two different destinations. In such a case, each destination would be examined independantly.
We are implementing a Dark Fiber Solution for all of our 4 sites to provide high bandwidth and private WAN Link. The 4 sites will have Ring Connection to each other which will be provided by the Service Provider. We are implementing BGP for the first time in the network.
The plan is to have the 4 sites in the same ASN number and I would run iBGP across the sites to create a full mesh connection.i would create a Layer 3 Interface on each Cisco 9300 switch in the same subnet.
My concern is routing to the ISP. if I want 2 of my sites to be Primary and Secondary Hub for routing to the ISP.I dont want to load balance Routes due asymmetric routing scenario.
What would be considered as the Best practice for routing to ISP in this case?
First of all, in such a scenario, I would either use an IGP between all sites (for faster convergence) or make each of the four sites a different AS and run eBGP between the sites, and an IGP within each site. Remember that typically, within an AS, it is the IGP that is used for routing, while iBGP is typically used to gather information about prefixes internal to an AS. These prefixes are then shared via eBGP between ASes.
Now having said all of that, it is possible to use iBGP within an AS to route traffic, but because it is generally slow to converge, it is not recommended. Now you may have a specific reason behind your particular design, and that’s fine.
If we were to keep your topology and design, how can you use iBGP to make sure that load balancing does not occur so that you avoid asymmetric routing? Well, if you use iBGP then you have nothing to worry about. BGP by default does not perform load balancing. It will always choose a single path to reach the destination. Unlike OSPF and EIGRP which both do equal cost load balancing if the metric for two or more paths is exactly the same.
So if you configure your BGP attributes so that the primary ISP is chosen as the best path, you have nothing to worry about when it comes to load balancing.