[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: Lambda LSP establishment

To: 'Heiles Juergen ' <Juergen.Heiles@icn.siemens.de>, ''John Drake' ' <jdrake@calient.net>, 'Maarten Vissers ' <mvissers@lucent.com>, 'manoj juneja ' <manojkumarjuneja@hotmail.com>
Subject: RE: Lambda LSP establishment
From: "Mannie, Eric" <Eric.Mannie@ebone.com>
Date: Thu, 13 Dec 2001 18:56:31 +0100
Cc: "'ccamp@ops.ietf.org '" <ccamp@ops.ietf.org>
Hi All,

A few comments,

>In MPLS the FA establishes "virtual" layer networks. In GMPLS the layer
networks already exist (SDH HO/LO-VCs, RS, MS, G.709 ODUs, OCh, OMS, OTS). A
FA basically corresponds to a real trail in a circuit switched network.

Eric: ...and this trail is setup dynamically using GMPLS. That's the
interest. One can think of SDH/SONET FAs as shortcuts or express-routes.

>One special thing about the SDH label is, that it includes a hierarchy as
it identifies the lower order VC in a higher order VC in a STM-N signal.
Note that this is somehow in contradiction with a statement in the GMPLS
signaling document which says 

>"A Generalized Label only carries a single level of label, i.e., it is 
   non-hierarchical.  When multiple levels of label (LSPs within LSPs) 
   are required, each LSP must be established separately, see [MPLS- 
   HIERARCHY]." 

No, hierarchy of labels is not hierarchy of layers. And even one should have
a common definition of what means a "layer". This text says that we cannot
have multiple embedded labels, e.g. one label containting a label stack.
This is a signaling stuff that doesn't imply anything on the transport
plane.

Now, a label identifies ONE LSP at an interface. This LSP can be HO or LO in
SDH. The label can be fully specified or partially specified, depending on
the context where the LSP is established.

E.g. when an LSP is established over a FA, the highest part of the LSP label
is not relevant. When the FA LSP is established the lowest part of the FA
LSP label is not relevant.

If a low order LSP is established without any higher order FA, the label is
fully specified. As you said the link is indeed the "ultimate FA" in that
case. Except that from the routing point of view a link and an FA are two
different things. There is a routing adjacency "over" a link but not over a
FA.

The SDH/SONET label just includes what you need to include in each scenario.

Moreover, the label MUST be interpreted according to the type of interface
for which it is used. It is possible to code two labels having the same
value but a complete different meaning. Labels are context sensitive of
course.

For instance, an LSP over an STM-0 interface or over a FA will have the
highest part set to zero, and could possibly have the same lowest part. You
cannot understand what means the label without knowning the detail of the
interface. In that case, the interface is either an STM-0 interface or a
VC-3 FA. And of course that FA and STM-0 interface are two different things,
but from the LSP point of view, in both cases what it wants is a VC-3 in
which it can be multiplexed.

Hope this helps.

I start to wonder if we should have an informational draft describing
different scenarios of SDH/SONET LSP establishment. That could complement
the SDH/SONET signaling drafts and the GMPLS architecture. That could solve
many terminology and modeling issues.

Kind regards,

Eric

-----Original Message-----
From: Heiles Juergen
To: 'John Drake'; Maarten Vissers; manoj juneja
Cc: ccamp@ops.ietf.org
Sent: 12/13/01 11:15 AM
Subject: RE: Lambda LSP establishment

Let me express my understanding of FAs in MPLS/GMPLS, please correct me
if I am wrong. I have also some questions related to the label
generation and interactions between layer networks.

The label in MPLS is local to the link between two adjacent MPLS
switches, it indicates a LSP in this link. This link is therefore the
ultimate FA. An already established LSP between two MPLS switches, which
don't have to be adjacent, can be used to transport/tunnel other LSPs
between this two nodes. This already established LSP generates a virtual
adjacency between the two nodes, the FA. As several LSPs can use this FA
it is also a kind of virtual multiplexing.
If you compare it with a circuit switched network the FA is a server
layer trail that provides transport (a link connection) for one or more
client layer signals.
In MPLS the FA establishes "virtual" layer networks. In GMPLS the layer
networks already exist (SDH HO/LO-VCs, RS, MS, G.709 ODUs, OCh, OMS,
OTS). A FA basically corresponds to a real trail in a circuit switched
network.
In GMPLS the label is also local to the link between the two devices
that perform the switching. For example for SDH the VC-N is identified
by the STM-N link/port and the SUKLM number according to
draft-ietf-ccamp-gmpls-sonet-sdh-02.txt. The STM-N port is identified by
the interface ID (I am not sure on this) and SUKLM is the SDH label that
locates the VC within the STM-N.
From just the SDH viewpoint the STM-N connection is the ultimate server
trail/FA. However the STM-N signal could be transported over a Optical
Channel or G.709 ODU. The OCh or ODU can already start in the equipment
that performs the VC-N switching. So the STM-N signal is not the port,
but the WDM signal is the port and the STM-N signal makes use of a OCh
or ODU server layer trail. This server layer trail can be established
via management or using GMPLS. 
Can someone explain how a label for a VC-4 is generated in this case
(pre-established OCh or ODU trail via management or setup using GMPLS).
This interaction between the different technologies/labels need in my
view some further explanation.

One special thing about the SDH label is, that it includes a hierarchy
as it identifies the lower order VC in a higher order VC in a STM-N
signal. Note that this is somehow in contradiction with a statement in
the GMPLS signaling document which says 
"A Generalized Label only carries a single level of label, i.e., it is
   non-hierarchical.  When multiple levels of label (LSPs within LSPs)
   are required, each LSP must be established separately, see [MPLS-
   HIERARCHY]."
Following this statement each layer should have its own label
independent of server layers.
The combination of hierarchies in labels is based on technology (e.g.
SDH, Sonet, G.709 ODU) but has in my view no real technical reason. The
SDH label for example fits to standard STM-N signals, but not to sub-STM
signals.
For the SDH the full SUKLM number is used if a e.g. VC-12 is located in
relation to a STM-N interface. If it is located in relation to a VC-4
(the VC-4 is in this case a FA) SUK are set to 0. For me it is not clear
in which case I use the first and in which case I use the second case as
a VC-4 trail is always needed for a VC-12 connection. This VC-4 could be
established using management or GMPLS. However it should have no
influence on the label. 
Some more information is needed in my view in this area.


Juergen
 

> -----Original Message-----
> From: John Drake [mailto:jdrake@calient.net]
> Sent: Wednesday, December 12, 2001 6:54 AM
> To: Maarten Vissers; manoj juneja
> Cc: ccamp@ops.ietf.org
> Subject: RE: Lambda LSP establishment
> 
> 
> fortunately, this is just your opinion
> 
> -----Original Message-----
> From: Maarten Vissers [mailto:mvissers@lucent.com]
> Sent: Tuesday, December 11, 2001 9:31 PM
> To: manoj juneja
> Cc: ccamp@ops.ietf.org
> Subject: Re: Lambda LSP establishment
> 
> 
> Manoj,
> 
> Forget the FA stuff, it is not appropriate in circuit 
> networks. It only
> applies
> to MPLS. We should remove it when it is used in relation with PDH,
> SDH/SONET,
> OTN and pre-OTN. The text in sdh-sonet draft should state 
> that if there is a
> LOVC link (IETF: link bundle/TE link) then the LOVC signals 
> use a label with
> "00KLM".
> 
> Regards,
> 
> Maarten
> 
> manoj juneja wrote:
> > 
> > Hi Marteen,
> >             The concept of FA is mentioned for SDH/SONET in 
> gmpls-sdh
> > -sonet draft. It says that if u have a HOVC trail as a FA 
> (advertized
> > as a link) then u can allocate the lower level signals in 
> it by making
> > the higher bits of label as 0s (i.e. S and U}. This is fine 
> for the same
> > technology. What about the case where the TDM LSP has to be tunneled
> through
> > the Lambda LSP ? What will be the form of label (i.e. 
> {SUKLM} or lambda
> > etc.) ?
> > 
> > Regards,
> > manoj.
> > 
> > >From: Maarten Vissers <mvissers@lucent.com>
> > >To: manoj juneja <manojkumarjuneja@hotmail.com>
> > >CC: ccamp@ops.ietf.org
> > >Subject: Re: Lambda LSP establishment
> > >Date: Tue, 11 Dec 2001 15:56:41 +0100
> > >
> > >Manoj,
> > >
> > >You refer to one wavelength to be available between A-E without
> wavelength
> > >conversion capability. This suggests that you operate at 
> the OCh layer
> > >network.
> > >Then you specify the capacity of the 4 OCh link 
> connections (A-B, B-C,
> C-D,
> > >D-E)
> > >to be "C". Say that C is about 10 Gbit/s. You then assume 
> that there is a
> > >request for an OCh signal with capacity C/4 (e.g. 2.5 
> Gbit/s) between C
> and
> > >E.
> > >The result is that the OCh link connections C-D and D-E 
> are transporting
> > >the OCh
> > >signal (of e.g. 2.5G). These OCh link connections are now 
> in service and
> > >not
> > >longer available to an other OCh connection request. I.e. 
> a request for
> an
> > >OCh
> > >connection between A and E will be rejected.
> > >
> > >FAs are not applicable in the circuit layers. IF there is 
> a trail in
> server
> > >layer X, then there is a link in its client layer Y. X and 
> Y are thus
> > >different
> > >layer networks and signals.
> > >
> > >If C-E is a "FA", then in an OTN the C-E connection would 
> be an OCh trail
> > >supporting an ODUk (k=1 if OCh is 2G5) link with a single link
> connection.
> > >
> > >Note a FA in MPLS creates essentially a MPLS sublayer 
> network. Such is
> not
> > >possible in the SDH/SONET, OTN, PDH or ATM technologies.
> > >
> > >Regards,
> > >
> > >Maarten
> > >
> > >OCh link connections
> > >
> > >manoj juneja wrote:
> > > >
> > > > Hi All,
> > > >         If I have 5 nodes A, B, C, D and E connected as 
> shown. Assume
> > > > that only one wavelength is available on the path A to E (no
> wavelength
> > > > conversion capability is there on the complete path). 
> Let the capacity
> > > > of the wavelength be C. Further assume a request arrives for
> connection
> > > > from node C to E for a line capacity of C/4. This 
> request will be
> > > > successful as we have available wavelength. Now If 
> another request
> > > > comes at node A to establish another connection from 
> node A to node E
> > > > via nodes {A,B,C,D,E} for a line capacity of C/4. 
> Should this request
> > > > be successful as we have already allocated the wavelength ?
> > > >
> > > > If the previous connection from C to E of capacity C/4 had been
> > > > advertised as a FA, in that case will the IInd request succeed ?
> > > >
> > > > If the previous connection from C to E of capacity C/4 
> had not been
> > > > advertised as FA then what will be the fate of IInd connection ?
> > > >
> > > >        A <--> B <---> C <----> D <---> <----> E
> > > >
> > > > Regards,
> > > > manoj.
> > > >
> > > > 
> _________________________________________________________________
> > > > Get your FREE download of MSN Explorer at
> > >http://explorer.msn.com/intl.asp
> > ><< mvissers.vcf >>
> > 
> > _________________________________________________________________
> > Send and receive Hotmail on your mobile device: 
http://mobile.msn.com
Prev by Date: RE: Moving right along ... generalized-signaling-06
Next by Date: RE: Lambda LSP establishment
Prev by thread: RE: Lambda LSP establishment
Next by thread: RE: Lambda LSP establishment
Index(es):
- Date
- Thread