Multiprotocol Label Switching (mpls)

Adrian Farrel | 5 Sep 2006 00:00

Re: WG Last Call on draft-ietf-mpls-number-0-bw-te-lsps-02.txt

Hi,

A quick review of this I-D leaves me puzzled by the motivation. I don't 
think it is enough to say "I want to know how many unconstrained TE LSPs 
traverse every link in the network"; we must have an idea of how that 
information will be used.

Now the I-D says:

... in the Abstract
   There are various
   circumstances (for example in order to load balance unconstrained TE
   Label Switched Path (LSP) across a set of equal cost paths) where it
   would be useful to also advertise the number of unconstrained Traffic
   Engineering Label Switched Path(s) (TE LSP) signalled across a link.

... and in the Introduction
   If the
   number of unconstrained TE LSPs traversing each link in the network
   is known, various algorithms can be designed so as to efficiently
   load balance the traffic carried onto such unconstrained TE LSPs.

...but this seems to me to be an supported assertion. I guess you could cite 
a reference so you don't need to prove the algorithms here. But it seems to 
me that the count of such LSPs is a very poor measure indeed. The only way 
you could use it would be if you could make some fairly tight statistical 
assumptions about the traffic on each LSP, or at least the statistically 
aggregate traffic on a number of LSPs.

Surely it would be better for the LSRs to report link usage rather than the

(Continue reading)

headers

Dimitri.Papadimitriou | 8 Sep 2006 12:07

Re: WG Last Call on draft-ietf-mpls-number-0-bw-te-lsps-02.txt

all - a couple of comments on this document

o) status intended for info or std track ?

o) end of section 1 - what about LSPs not instantiated via signaling ? are 
they also incorporated in the LSP counter ?

o) section 2 - "Unconstrained TE LSP" why only bandwidth enters in the 
definition of "unconstrained" or does it become now an new equation 
constraint = bandwidth ?

o) as the document states "There are various circumstances (detailed 
below) where it would be useful to also advertise the number of 
unconstrained Traffic Engineering Label Switch Path(s) (TE LSP)."

what is the use of the other TLVs describing unreserved/reservable 
bandwidth in these conditions ? what about overprovisioning (wrt #LSPs) ?

o) as the document involves a procedure "An implementation may decide to 
implement a dual-thresholds mechanism to govern the origination of updated 
OSPF LSA or ISIS LSP."

- threshold on which basis ? -

o) more fundamental - bandwidth gives an implicit limitation on the number 
of signaling states an interface could receive - the mechanism disappears 
here, counting is started from 0 if my reading is correct (the draft does 
not even clearly mention how counting is performed and translated into the 
field assigned in the newly defined TLV) 
what is the AC mechanism provided such as to prevent crashing the

(Continue reading)