BGP (I) - Emulating MPLS Service

Problem Description

I was working on a project require me to emulate a customer MPLS enviornment, so that we can do furthter networking architecture migration and recontruction. In brief, below is a simple emulated MPLS backbone network running iBGP with ASN 65000. And CE router also has its own BGP reign with ASN 65001, which means it runs eBGP with the service provider.

CE1 <--eBGP--> PE1 <--iBGP--> PE2 <--eBGP--> CE2

So here is a problem: when I show ip bgp in four machine, I certainly cannot see the BGP routing tables in both CE1/CE2 router, while there exist the BGP routing flows in PE routers.

What reason was that PE can recieve the BGP advertising information, which CE can’t?

How to Solve it

My co-worker and also probably the best mentor, Wei, told me the answer lie behind in how BGP is designed to prevent routing loops by rejecting routes that contain the local AS number in the AS path as default behavior. This mechanism helps ensure that a route is not advertised back into the AS from which it originated, which could cause routing loops and instability in the network.

Thus the solution is pretty simple, by adding a function allowas-in, the CE routers will then be able to recieve routes of ASN 65001 from PE routers.

From:

router bgp 65001
 bgp log-neighbor-changes
 network 1.1.1.5 mask 255.255.255.255
 redistribute connected
 neighbor 192.168.1.1 remote-as 65000

To:

router bgp 65001
 bgp log-neighbor-changes
 network 1.1.1.5 mask 255.255.255.255
 redistribute connected
 neighbor 192.168.1.1 remote-as 65000
 neighbor 192.168.1.1 allowas-in

As a result, our CE routers have obtain the BGP routes to the other site.

Backgroud Knowledge - CCNP perspective

For every network engineer with Cisco training progress, they must have this table in mind. Although it’s quite easy to forget for most people, the idea of the straight forward table really help us keep that concept in mind.

BGP Route Selection
1. Prefer the path with the highest WEIGHT.
2. Prefer the path with the highest LOCAL_PREF.
3. Prefer the path that was locally originated via a network or aggregate BGP subcommand or through redistribution from an IGP.
4. Prefer the path with the shortest AS_PATH.
5. Prefer the path with the lowest origin type.
6. Prefer the path with the lowest multi-exit discriminator (MED).
7. Prefer eBGP over iBGP paths.
8. Prefer the path with the lowest IGP metric to the BGP next hop.
9. Determine if multiple paths require installation in the routing table for BGP Multipath.
10. When both paths are external, prefer the path that was received first (the oldest one).
11. Prefer the route that comes from the BGP router with the lowest router ID.
12. If the originator or router ID is the same for multiple paths, prefer the path with the minimum cluster list length.
13. Prefer the path that comes from the lowest neighbor address.

reference: https://www.cisco.com/c/en/us/support/docs/ip/border-gateway-protocol-bgp/13753-25.html

Backgroud Knowledge - CS perspective

As I just said, the prolix table always lead to oblivion. Therefore I tend to think it in a different way. From solid four years of computer science education and training process, I prefered to associate dynamic routing protocol to CS fundamental algorithm(yes, those algorithms implemeneted to graph theory):

1. distance-vector
   • Algorithm: Bellman-Ford algorithm.
   • Route Calculation: Based on the distance to the destination(vector).
   • Information Sharing: Each router shares its entire routing table with its immediate neighbors at regular intervals(aka. gossip algorithm).
   • Convergence Time: Generally slower; prone to issues like counting to infinity.
   • Complexity: Simpler to implement and understand.
   • Scalability: Less scalable; suitable for smaller networks.
   • Examples: RIP, IGRP.
2. link-state
   • Algorithm: Dijkstra's algorithm.
   • Route Calculation: Each router constructs a complete map (link-state database) of the network topology and calculates the shortest path.
   • Information Sharing: Routers exchange link-state advertisements (LSAs) with all other routers in the network.
   • Convergence Time: Faster; more efficient handling of topology changes.
   • Complexity: More complex to implement and requires more resources.
   • Scalability: More scalable; suitable for larger networks.
   • Examples: OSPF(intra area), IS-IS.
3. hybrid
   • Algorithm: Combination of distance-vector and link-state techniques.
   • Route Calculation: Uses distance-vector techniques for quick updates and link-state techniques for accurate information. Examples: EIGRP.
4. path-vector
   • Algorithm: Distance-vector algorithm extension.
   • Route Calculation: Based on the path information (list of ASes) that a route has traversed.
   • Information Sharing: Each router shares the path information with its peers, including the distance and the entire AS path.
   • Convergence Time: Depends on path propagation; can be slower in very large networks.
   • Complexity: High flexibility and control over route selection using various attributes.
   • Scalability: Highly scalable; suitable for inter-domain routing.
   • Examples: BGP.

According to above catagorized algorithm types, we can determine that BGP is one implementation of path vector algorithm. And the paragraph below is aim at explaining BGP with relative point of view of graph theory:

So, when you show ip bgp on a router, you will get informations to each network, like NextHop, LocPref, Weight, and Path. But what’s that suppose to mean?

Metric

The MED (Multi-Exit Discriminator) is a BGP attribute used to suggest a preferred entry point into your AS when multiple paths exist from the same neighboring AS. Think of it as your neighbor politely saying:

“If you have several ways to reach me, please take this one!”

You might ask:

“Why this path?”

And your neighbor replies:

“Because this link has 10 Gbps bandwidth. The other one barely manages 1 Gbps.”

Notes: Lower MED = More preferred

Local Preference

Local Preference is for your local router’s internal compass. It’s how routers inside your AS agree on the best way out to the internet when multiple exit points exist.

Imagine you’re a captain managing traffic inside your organization. You walk into the chief room and declare on a advertisor:

“Everyone, if you see a path through ISP-A, take it! That’s our main way.”

Routers across your AS nod and say:

“Aye, Captain. Local_Pref 200 for ISP-A. I will make sure all traffic go through me will be sent to ISP-A.”

Meanwhile, ISP-B serves as a Local_Pref of 100 — not terrible, but not the first pick either.

AS_Path

AS_Path is like the route record stamped onto a passport. Every time a BGP UPDATE passes through an AS, it appends its ASN to the list — kind of like leaving footprints across the internet. In the internet, eBGP network are too gigantic to manage. Therefore, the best strategy to route traffic in this opaque environment is to make sure sending packet through the path as short as possible.

When a router receives two routes to the same destination, it asks itself:

“Hmm… one path passed through 3 ASes, and the other only went through 1. I’ll take the shorter trip.”

Notes: Shorter AS Path = more preferred

It’s also a loop prevention mechanism — if your ASN appears in the path, the route is rejected

You can manipulate AS Path deliberately using AS Path Prepending

Bonus Section: Local Preference vs. AS Path — Which is Better for Controlling Outbound Traffic?

When designing outbound routing policies, both Local Preference and AS Path Prepending can influence path selection — but they serve different purposes and are best used in different scenarios.

• Use Local Preference when:

  You have full control on all routers within your AS. You want a monothlic and reliable mechanism to decide preferable egress point. So that you can make entire egress traffic from your AS to goto ISP-A instead of ISP-B.

• Use AS_Path Prepending when:

  It’s not your routers even not your AS, and you’re not allow to change upstream ISP routing policy. You want to influence how traffic flow to reach you by unprefering certain inbound paths without blocking them. So that you can make entire ingress traffic mainly from ISP-A and then from ISP-B.