The MPLS WG Archive

[Shahram Davari <Shahram_Davari@pmc-sierra.com>]

Thanks Curtis,

I agree with your conclusion.

-Shahram

> -----Original Message-----
> From: Curtis Villamizar [mailto:curtis@fictitious.org]
> Sent: Tuesday, December 10, 2002 1:01 PM
> To: Shahram Davari
> Cc: 'curtis@fictitious.org'; Ajay Simha; mpls@UU.NET
> Subject: Re: Record Object in LSP Ping 
> 
> 
> 
> In message 
> <4B6D09F3B826D411A67300D0B706EFDEB03B72@nt-exch-yow.pmc-sierra.bc.ca
> >, Shahram Davari writes:
> > I fail to understand how adding a TLV to the ping affects 
> > forwarding ASICs!
> > 
> > -Shahram
> 
> 
> Shahram,
> 
> There are two modes:
> 
>   1.  ping mode
> 
>   2.  traceroute mode
> 
> Since traceroute mode already yields the full path, the record route,
> analogous to IP record route option, would be applicable to ping mode
> or to traceroute once TTL was greater than 1.
> 
> In traceroute mode, a control plane trace is performed and confirmed
> by incrementing TTL and checking to see that the router returning the
> next reply was the one expected, thereby verifying each forwarding hop
> separately.
> 
> In ping mode and when traceroute TTL>1 , the forwarding is being
> exercised.  The only way a record route can be supported is to modify
> the forwarding such that the bottom of the stack is checked, the
> payload determined to be mpls-ping, the packet extracted from the
> forwarding stream, the record route hop recorded, and the packet put
> back into the forwarding stream.  This arguably is no longer the
> normal packet forwarding so not only would we (everyone) have to
> change hardware, but we defeat a prime purpose of ping mode (a
> forwarding based continuity check).
> 
> Curtis
> 
> 
> > > -----Original Message-----
> > > From: Curtis Villamizar [mailto:curtis@fictitious.org]
> > > Sent: Tuesday, December 10, 2002 12:19 PM
> > > To: Ajay Simha
> > > Cc: mpls@UU.NET
> > > Subject: Re: Record Object in LSP Ping 
> > > 
> > > 
> > > 
> > > In message <20021210114444.GL677@cisco.com>, Ajay Simha writes:
> > > > Hi,
> > > > 
> > > > Since the LSP Ping is meant for fault detection, does 
> > > anyone see any value
> > > > in adding a record route/label TLV to this?
> > > 
> > > This is infeasible without changing forwarding hardware so "no".
> > > 
> > > > Each node can add (if requested) the label that the packet 
> > > came in with,
> > > > the label that this node imposed/swapped and it's LSR-ID.
> > > 
> > > All you need is new forwarding ASICs.
> > > 
> > > > Thanks,
> > > > 
> > > > -ajay
> > > 
> > > Curtis
> > > 
> > 
>