[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: Sonet Ring provisioning
Hi,
If a path that is set up via GMPLS signaling traverses a SONET ring, and the
SONET ring is managed via an EMS as Greg suggests, how do the two protection
domains interact? The two protection domains I am thinking of are as
follows:
1) The Ring protection domain - failures in the ring itself are protected,
by ring-specific mechanisms
2) The end-to-end protection domain - failures anywhere along the path are
protected, presumably by failing over to a diversely routed path
J------K
/ \
A ------B------H------I L-------O-------Y--------Z
\ \ / /
\ N------M F------G
\ /
C---------------D---------E
Endpoints A and Z have an LSP (an end-to-end SONET path) switched across the
ring (I-J-K-L-M-N). LSRs B and Y provide a redundant LSP across the path
C-D-E. Presumably, the nodes B and Y provide a path monitoring function,
similar to UPSR, that they can use to initiate a switch to the protect LSP.
If a failure happens in the ring, it will be recovered from, via
line-switching or ring-switching inside the ring. But the path monitoring
function at nodes B and Y will invoke an LSP protection switch. This is
marginally bad, because it extends the outage by some number of
milliseconds, and it complicates the revert later.
If the path monitoring function has a holddown timer, such that it does not
invoke a switch until 50 ms of outage, then failures at H or O will not be
reacted to as quickly as possible.
Instead of performing path monitoring to trigger an LSP-switch, B and Y
could perform line monitoring, and only switch if failures occur on their
local segments? But this seems like it opens up a class of failures which
would not be recovered from.
So it seems to me that the mixing of protection schemes creates a small
problem, and I wonder what people think about this?
Nik
> -----Original Message-----
> From: Zhi-Wei Lin [mailto:zwlin@lucent.com]
> Sent: Thursday, May 30, 2002 8:20 AM
> To: R. Muralidharan
> Cc: Bernstein, Greg; 'Manoj Agiwal'; 'Ccamp (E-mail); mpls@UU. NET
> (E-mail)
> Subject: Re: Sonet Ring provisioning
>
>
> Hi all,
>
> Seems we have mixed up a "ring topology" from "ring
> protection mechanism".
>
> * In a general ring topology, *any* recovery mechanism
> may be used.
> This may be the underlying transport's BLSR/MSSPRING or
> UPSR/SNCPRING, or it may be using control plane to set
> up 1+1 or 1:1.
> * In a BLSR/MSSPRING ring, the type of protection has already been
> decided (i.e., BLSR *is* the name of the protection mechanism).
> Similarly for UPSR/SNCPRING.
>
> Thus if we're talking about BLSR/UPSR rings, then the
> question is: how
> does the control plane protocols interact (do they interact?)
> with the
> underlying BLSR/UPSR, e.g., does the control plane simply treat the
> *entire* ring as a single node and thus interfaces to
> existing EMSs to
> ask for a sub-network connection across the "node", or does
> the control
> plane actually see all nodes of the ring and ask individual
> nodes for an
> STS-1/VC-3 connection (but note it only asks for the normal
> channel i.e.
> one connection, since the protection channel comes by default -- the
> service is either protected or unprotected but one connection
> in either
> case)...
>
> Zhi
>
>
> R. Muralidharan wrote:
>
> >Hi All,
> >
> >My understanding is as follows:
> >
> > BLSR or UPSR rings may be already provisioned in the
> SONET network. The
> >task using GMPLS is to set up an end to end virtual path, may be
> >encompassing multiple SONET rings. This virtual path set up
> is dynamic in
> >nature and the life of the path may be only for a fixed
> interval, after
> >which it would be tore down. When one wants to set up a path
> through a SONET
> >ring, one may have to specify whether one wants a 1+1
> protection path or a
> >1:N protection path or just an unprotected path. Based on
> this specification
> >the GMPLS can discover and set up an appropriate path in a
> BLSR or an UPSR
> >ring as the case may be and hence the cost would be optimum.
> ( assuming that
> >a 1+1 path would cost more than an unprotected path). As
> Greg pointed out, I
> >also believe that the BLSR and UPSR ring configurations
> would already have
> >been set up by associated NMS or EMS and advertised. The
> granularity of the
> >path that can be set up using GMPLS could a VT1.5 or
> multiples of them ( VT
> >Path) or STS-1s (STS Path).
> >
> >I have used the words "path" and "virtual path" in generic
> terms and not in
> >the SONET or SDH domain definitions.
> >
> >Does the above make sense ?
> >regards,
> >murali
> >OSS Systems India
> >
> >----- Original Message -----
> >From: "Bernstein, Greg" <GregB@ciena.com>
> >To: "'Manoj Agiwal'" <ManojA@netbrahma.com>; "'Ccamp (E-mail)"
> ><ccamp@ops.ietf.org>
> >Cc: "mpls@UU. NET (E-mail)" <mpls@UU.NET>
> >Sent: Wednesday, May 29, 2002 10:28 PM
> >Subject: RE: Sonet Ring provisioning
> >
> >
> >
> >
> >>Hi Manoj, the UPSR and BLSR cases are very different. I'm
> assuming that
> >>
> >>
> >you
> >
> >
> >>are setting up SONET paths such as STS-1, STS-3c, ...
> STS-192c or their
> >>
> >>
> >SDH
> >
> >
> >>equivalent.
> >>Then
> >>(1) In the BLSR case the protection is at the SONET line
> layer, i.e.,
> >>
> >>
> >below
> >
> >
> >>the layer of the connection that you are setting
> up(SONET/SDH are layered
> >>networks). In this case no special action needs to be
> taken unless of
> >>course the entity requesting the connection asked for "enhanced"
> >>
> >>
> >protection
> >
> >
> >>in the setup request.
> >>
> >>(2) A UPSR works at the path layer, i.e., like a path layer
> 1+1, with the
> >>redundant paths taking different routes around the ring.
> Hence you are
> >>actually setting up two connections that have a special
> relationship with
> >>each other. This would have to be indicated somehow (tunnel ID or
> >>something). Some UPSRs may put constraints on the timeslots
> used too. One
> >>meta question is why bother signaling around a UPSR versus
> talking to a
> >>
> >>
> >UPSR
> >
> >
> >>as a separate "protection domain"? There is no way mesh
> restoration could
> >>come close to a UPSR's restoration speed (it just selects
> the better of
> >>
> >>
> >two
> >
> >
> >>signals). Hence I don't understand the benefit signaling
> around the ring
> >>
> >>
> >in
> >
> >
> >>this case, all vendors have methods for setting up their
> UPSRs via EMS's
> >>
> >>
> >why
> >
> >
> >>not just interface to those rather than to the individual
> ring elements...
> >>
> >>Greg B.
> >>
> >>***********************************
> >>Dr. Greg M. Bernstein
> >>Senior Technology Director, Ciena Corporation
> >>
> >>
> >>-----Original Message-----
> >>From: Manoj Agiwal [mailto:ManojA@netbrahma.com]
> >>Sent: Wednesday, May 29, 2002 2:01 AM
> >>To: 'Ccamp (E-mail)
> >>Cc: mpls@UU. NET (E-mail)
> >>Subject: Sonet Ring provisioning
> >>
> >>
> >>Hi ,
> >> Do we use GMPLS signaling protocol for configuring
> Sonet Rings (
> >>
> >>
> >UPSR
> >
> >
> >>/ BLSR ( 2 fiber/4
> >> fiber ) ?
> >>
> >> I was going through certain white papers where there
> was a mention
> >>that GMPLS is used as
> >> a signaling protocol for provisioning mesh topologies only
> >>.Traditional Sonet rings will be
> >> replaced by mesh topologies in future .
> >>
> >> But there is a section 12.( LSP protection and
> restoration) in GMPLS
> >>Architecture which says
> >> that " Both mesh and ring like topologies are supported "
> >>
> >> How do we provision nodes in Sonet ring using GMPLS ?
> >> Or GMPLS has been designed to provision mesh topologies only.
> >>
> >>Regards ,
> >>Manoj
> >>
> >>
> >>
> >>
> >>
> >>
> >>
> >
> >
> >
>
>