Re: label space

Hi Rober

Could we state the following :

- For label distribution with direct/immediate neighbor :

The box could use per interface or per plateform

-For indirect neighbor (sevral hop away and distribution say by BGP) :

It is mandatory that the labels to be from the per plateform pool .

- reg the range of label per pool .

The per-interface pool say for if1 could intersect with any other

per-interface pool (if2 ...ifn) . .

The per-plateform pool should not intersect with any per-interface pool .

Reg :

Quote

>* For packet with labels comming from a SP backbone to the PE we don't

>check the VPN label - to be honest I am not even sure what should the

>check look like in either case per platform or per interface label spac

Unquote

guess that the BGP/MPLS VPN draft touch this point in section 13 page 33

Quote

- the packet's top label has a label value which the receiving
       system has distributed to a system beyond that neighbor (i.e.,
       when it is known that the path from the system to which the label
       was distributed to the receiving system may be via that
       neighbor).
Unquote

I understood that the PE should check that the labels advertised for

the VRF routes should only be accepted from interfaces which could

be in the forwarding path to other PE's .

Fot the case of full MP-iBGP mesh the PE will know all other PE's

address and by checking the LIB the PE could find out all eligibe interfaces.

ie: say PE1 advertise for certain VRF route a label (L) to PE2 .

PE1 check the LIB for all interface where it recieve a label for PE2 ,say

the list is {s1,s2} . Label (L) will only be accepted from s1 or s2 .

In case of RR , the PE will not be aware reg other PE's address so

the rule is compromised to accept label (L) from any global interface (not associated with any VRF) .

Could U pls comments.

Brgds

>From: Robert Raszuk

>Reply-To: raszuk@cisco.com

>To: Juha Heinanen

>CC: Rajiv Asati , mpls-ops@mplsrc.com, Eric Rosen

>Subject: Re: [MPLS-OPS]: label space

>Date: Mon, 29 Jul 2002 16:42:34 +0200

>Hi Juha,

> > in the "per platform case", how does cisco prevent packet misdelivery

> > due to incorrect incoming label?

>First I don't see a relation of your first part of the question the the

>second one. You may have multiple interfaces to the core and per

>interface label space will not buy you much as far as the VPN label

>processing is of concern.

>Now to answer the second part we do the following:

>* For packet with labels comming from CE (Carrier's Carrier case) we do

>check if packet is comming with a label advertised to a given site by

>PE. If this check fails the packet is discarded.

>* For packet with labels comming from a SP backbone to the PE we don't

>check the VPN label - to be honest I am not even sure what should the

>check look like in either case per platform or per interface label space

>....

>Rgs,

>R.

> > Juha Heinanen wrote:

> >

> > Rajiv Asati writes:

> >

> > > It depends.

> > > The router can do both.... "per platform label space" for Frame Mode MPLS,

> > > whereas "per interface" for Cell Mode MPLS.

> > > Please see the details here:

> >

> > in the "per platform case", how does cisco prevent packet misdelivery

> > due to incorrect incoming label?

> >

> > -- juha

>-------

>The MPLS-OPS Mailing List

>Subscribe/Unsubscribe: http://www.mplsrc.com/mplsops.shtml

>Archive: http://www.mplsrc.com/mpls-ops_archive.shtml

The MPLS-OPS Archive

Re: label space