[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Question regarding bi-directional connection set up



Fong,

I totally agree with you that bi-directional LSP for SONET/SDH network is a
must. The issue is how to support it.

It's useful to see how SONET/SDH NEs' ports are identified first. It is
different from IP router, which always use bi-directional interface ID (no
matter IP addressed or unnumbered). Some vendors identify the physical port
bi-directionally. (Port ID xxx actually indicate two physical ports, one IN, one
OUT). Other vendors identify the physical port uni-directionally. (transmitter
and receiver have different Port ID)

This has two implications:

(1) Label has directionality
(2) For SONET NEs whose port are bi-directionally identified, no upstream label
is needed for bi-directional LSPs

Take a step further, Upstream label is not necessary at all. For optical
network, link connection creation direction is independent from the user traffic
flow direction (Here is where IP network is different from optical network).
When  A-----B want to set up a bi-directional link, either A or B can choose
which ports to use.  

Regarding whether or not port ID should be part of the label. I think it should.
Port ID can be in the ERO but I don't think that's the good place because:

(1) ERO is the result of source node source routing. It is in bundle ID
granularity. (a bundle ID is assigned to a collection of physicall ports that
share the same routing significance).

(2) PortID + channel/subchannel ID is a meaningful unit. Breaking them into
pieces doesn't make sense

All details can be found in 

http://www.ietf.org/internet-drafts/draft-xu-ccamp-gmpls-sig-reorg-00.txt

Regards,

Yangguang  



>  > Yangquang
>  >
>  > I read your draft which puts port id in lable object
>  > and probably upstream label as well.  Do you not consider
>  > allowing asymmetric link for a connection a problem,
>  > in particular for SONET/SDH ?
>  >
> I think the rational of providing bi-directional LSP
> really should be enhanced to not just an optimization,
> but a MUST for SONET/SDH etc.