RE: [PWE3] PWE3 question

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

Eric,

> -----Original Message-----
> From: Eric Rosen [mailto:erosen@cisco.com]
> Sent: Wednesday, May 15, 2002 11:57 AM
> To: Rutemiller, John
> Cc: 'pwe3@ietf.org'
> Subject: Re: [PWE3] PWE3 question 
> 
> 
> 
> John> Protocols should not dictate network architecture.
> 
> John> The choice of whether an operator desires to use a separate
> John> label inside the network for each psuedo-wire, or to use a
> John> label stack to improve scaling inside the network is a
> John> network design issue.
> 
> John> It is not a protocol design issue.
> 
> It is  possible to use  a two-layer stack  and still design a 
>  network which
> lays  down state  for each  PW in  the core  routers, and  
> hence is  just as
> unscalable  as an ATM  network.
  The  protocol design  does 
> not  dictate the
> network design. 
> 
> The primary  issue is, do we really  need a protocol which  
> is optimized for
> unscalable network designs (uhh, I mean for network designs 
> in which each PW
> has state in the core routers)?  Anyone who thinks so should 
> make a proposal
> and should  make it clear why  the use of one  rather than 
> two  labels is so
> important.

1) The use of 1 label reduces the number of label lookups that needs
to be supported at PEs. Same objective as PHP!
2) If the MPLS networks must always be designed scalable, then perhaps
MPLS standard should mandate hierarchical LSPs, because all single level
LSPs have the scalability issue that you are referring to.

> 
> The secondary issue is whether  any particular protocol 
> should be considered
> to be universally mandatory (for conforming products to 
> provide), or whether
> each can just be implemented within its own scope of applicability.

I think each is applicable to its scope.

> 
> There is also an issue as to whether there is some realistic 
> set of service
> characteristics which  cannot be provided  unless the core  
> routers maintain
> state per PW.

The issue is opposite: Is there a realistic set of service characteristics which
can be provided even when the core routers maintain state per PW? 

>  will be hard  to get consensus on a  "one 
> label" protocol
> unless one  can first get  consensus on this.   Since the use 
> of  two labels
> does  not rule  out the  creation of  per-pseudowire state  
> within  the core
> routers, I'm not sure why folks even want to fight this battle.

>
The issue is not the need to keep the state at the core,
rather it is why it MUST be precluded?

-Shahram

> 
> 
> 
> 
> 
> 
> 
> 
> _______________________________________________
> pwe3 mailing list
> pwe3@ietf.org
> https://www1.ietf.org/mailman/listinfo/pwe3
> 

_______________________________________________
pwe3 mailing list
pwe3@ietf.org
https://www1.ietf.org/mailman/listinfo/pwe3