[c-nsp] MPLS & Bandwidth Management

[Mark Tinka]

Hi all.

I have a lab scenario where a border and edge router are directly 
connected through a bandwidth management device. Further down, 
another edge router simulating another PoP is connected to the 
main edge router that simulates the main PoP.

This bandwidth manager manages packet bandwidth utilization by 
analyzing various attributes. Naturally, we analyze the source 
IP address and apply bandwidth management policies on that.

I've noticed for prefixes whose out-label isn't a 'Pop tag' 
to/from the main PoP edge router will not have their bandwidth 
policies applied to them as they try to exit the AS domain.

After disabling LDP on the border router only, said prefixes are 
now adhering to the bandwidth management policies on the 
bandwidth manager.

This looks like the bandwidth manager doesn't know how to handle 
the packets because it sees the label at the top of the packet, 
rather than the header with the source IP address.

It's likely someone else has come across this issue. I'd 
appreciate any feedback on how it was handled (I'm investigating 
whether my bandwidth manager vendor can support policies based 
on MPLS labels, but that would still present a challenge for 
customers as label allocation is only done for the BGP next-hop 
IP address, and not an end-customer prefix).

I'm still analyzing some more, but for now, it would appear this 
mostly affects return traffic from the border router toward the 
end edge router/prefix.

All help appreciated.

Cheers,

Mark.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 827 bytes
Desc: not available
Url : https://puck.nether.net/pipermail/cisco-nsp/attachments/20060921/9070837b/attachment.bin