Misc Questions

[joyride_us]

Hi,

now this is for real for me. a few questions then :

1- can I set the 'ip vrf forwarding...' on a sub-interface or do I have to have 1 vrf per physical interface

2- is there a way I can set route-targets per prefix in the same vrf ?

3- can I have only PE (no P) routers. I would think so but need to be 100% sure!

Thanks

[mheusinger · Accepted solution]

Hi,

A1 - you can use "ip vrf forwarding ..." on subinterfaces as well.

A2 - the feature you are looking for is called conditional advertisement. Example config:

ip vrf Test

rd 65000:1

export map RT4prefix

route-target import 65000:123

route-map RT4prefix permit 10

match ip address prefix-list MyNet1

set extcommunity rt 65000:1 65000:456

route-map RT4prefix permit 20

match ip address prefix-list MyNet2

set extcommunity rt 65000:2

ip prefix-list MyNet1 permit 10.1.1.0/24

ip prefix-list MyNet2 permit 192.168.0.0/16 le 32

A3 - You do not nssessarily need P routers. You can connect PEs back-to-back or even partially mesh only PE routers. They will perform label switching for transit traffic.

Hope this helps! Please rate all posts.

Regards, Martin

[joyride_us]

Thank you Martin! Exactly what I needed!

[joyride_us]

by the way Martin (since you seem to be around :)

I am seriously considering to implement MPLS in the internal (large) network of a company because of the flexibility to separate traffic flows with virtual routing table and flexibility to filter/distribute prefixes with the export-map/route-target.

Last, a still nice feature I would like to use is FRR with tunnels for fast routing failover between the PE's.

what do you think ? Does that sound an abuse of MPLS ?

Thanks!

[joyride_us]

by the way Martin (since you seem to be around :)

I am seriously considering to implement MPLS in the internal (large) network of a company because of the flexibility to separate traffic flows with virtual routing table and flexibility to filter/distribute prefixes with the export-map/route-target.

Last, a still nice feature I would like to use is FRR with tunnels for fast routing failover between the PE's.

what do you think ? Does that sound an abuse of MPLS ?

Thanks!

[olorunloba]

It does not sound like an abuse to me. Sounds like what MPLS is supposed to do. However, note that FRR is only available on the higher end routers like the GSRs

[joyride_us]

Good point with FRR...Need to re-think that aspect although it was not essential either.

I think those expensive routers one used to have to buy for MPLS has been a show-stopper for MPLS deployment. Now that that it is coming on lower-end models, I find the applications of MPLS attractive and affordable!

[mheusinger]

Hi,

well in the early days of MPLS it was developped for Service Providers. But nowadays it?s moving more and more into the enterprise area. Mainly because often enterprise IT departments find themselves in the role of a company internal SP. So lately there were a lot of "MPLS in the datacenter" articles, presentations and the like.

So use it, IF it solves your problems AND you know what you are doing. And get your staff (1st level!) properly trained.

Instead of FRR you could reduce your routing timers. With proper hardware in place convergence times can be reduced to a second or so, which is acceptable by practically every application (or user ;-)

Hope this helps! Please rate all posts.

Regards, Martin

[joyride_us]

Hmmmm. No, I will not start playing with routing timers :) I have some friends who have tried and the conclusion was : never again : something has to be wrong on one of the routers which we will not necessarily find out at once (like only under convergence period) and it will be hell to figure out what is going on (just before getting hung by the IT Director!)

:)

[olorunloba]

On which IGP was the problem? How was it implemented? I've used fast hellos with OSPF without any issues. I think this can be combined with LDP targetted hellos for fast convergence