The MPLS WG Archive

[David Allan <dallan@nortelnetworks.com>]

Hi Kireeti:

> > LSP-PING can also be used for RTT measurements, which when 
> performing 
> > the FEC stack check when processing a request message I 
> would consider 
> > to be somewhat imprecise....
> >
> > Perhaps the FEC stack and associated processing should be 
> optional if 
> > useful RTT measurements are to be performed?
> 
> RTT measurements are usually accurate to milliseconds or 
> tenths of msec.  Do you think that FEC processing will really 
> affect that?

Depends on the implementation and what else the LSR happens to be doing at
the time. The boilerplate in the BFD for LSP drafts discusses that PING does
require non-zero processing resources...ditto for LSR-SELF-TEST which makes
the FEC stack optional to reduce the load.

> Besides, doing RTT on the wrong ping packet because the FECs 
> were not processed seems worse than jittering the result by a 
> few tens of microseconds.

T'would be nice to assume that a single god packet can do everything at
once, but I'm personally rather circumspect. IMO one ping to check the FEC,
and one ping to do RTT is not an onerous burden...

Given that the LSR SELF TEST PDU exists at least partially to eliminate the
FEC stack, can this be reconciled. (e.g. make the TLV optional rather than
defining a new message type)... 

cheers
Dave