1dc545a23e89d4a6acfd4e6d28ccd619302c8a7d
howto/mpls.md
... | ... | @@ -0,0 +1,25 @@ |
1 | +mpls label switching is faster because it's a divide and conquer search in an ordered list, compared to routing, which is a longest prefix match search |
|
2 | + |
|
3 | +and doing just label switching, especially with multiple labels, have consequences like |
|
4 | + |
|
5 | +you can provide vpns, be that layer2 or layer3 on the same infra, we can source-route through arbitrary paths we want, and so on.... |
|
6 | + |
|
7 | +you can control visibility / reachability by route target export / imports, so you can hide various routes from specific endpoints, then they'll become unreachable just for them, basically rendering packet filtering unnecesary |
|
8 | + |
|
9 | +hiding service addresses (ip / mac) from the infra resulting in less resource needs: in the simplest mpls, you dont need bgp route table only where the packet enters the network |
|
10 | + |
|
11 | +you can hide your core from traceroute by disabling ip ttl protopagation |
|
12 | + |
|
13 | + |
|
14 | + |
|
15 | +participating networks: |
|
16 | + |
|
17 | +nop-mnt |
|
18 | + |
|
19 | + |
|
20 | + |
|
21 | +planned: |
|
22 | + |
|
23 | +C4TG1RL5-famfo |
|
24 | + |
|
25 | +Fortless |