Re: SE style in optical neyworks

[Maarten Vissers <mvissers@lucent.com>]

Sudheer,

In a layered network approach, a connection request for e.g. a VC-4 connection
will use the VC-4 layer network with its established VC-4 links (G.805
terminology) and VC-4 fabrics.

Some VC-4 links may have higher availability than others. This may be due to
(but is not at all limited to) supporting such VC-4 link by a protected server
layer trail (e.g. MS/Line protected, MS SPring/BLSR protected). The availability
of the link could be a parameter in the route selection process.

If the VC-4 connection is to be protected (i.e. requirement is to go via two
diverse routes), 1+1 VC-4 SNC protection is to be applied between e.g.
ingress/egress of the network independent of the individual availability figures
(and possible protection underneath) of the VC-4 links.

If there is just a requirement to meet a particular availability value for the
requested connection, then VC-4 links and VC-4 fabrics should be selected such
that the total availability figure doesn't exceed the requested value. Now it
may not be required to have complete diverse routed working and protection
connections; i.e. intermediate nodes may be common, which allows to use server
layer protection for some parts of the VC-4 connection.

Instead of using 1+1 VC-4 SNC protection, you can use VC-4 restoration. VC-4
restoration may require an alternative VC-4 connection to be reserved. In this
case once again the VC-4 alternative connection should be computed using the
VC-4 links and VC-4 fabrics. How these links are supported doesn't matter, as
long as they are disjoint from the VC-4 links used for the normal connection.
You may oversubscribe VC-4 links with alternate/backup VC-4 connections, as long
as a single fault in the network will not cause an overflow of the VC-4 link
while too many reserved VC-4 connections must be turned into actual connections.

Regards,

Maarten

Sudheer Dharanikota wrote:
> 
> Hi John:
> 
> John Ellson wrote:
> 
> > Sudheer Dharanikota wrote:
> > > Hi Gentelmen:
> > >
> > > I changed this list to camp, as it is more appropriate
> > > for this discussion.
> > >
> > > I would like to understand the following..
> > >
> > > Assumptions:
> > >
> > > - Segments of network are inherantly made protected.
> > >   For example, as suresh said, span/UPSR/BLSR etc protected.
> > >
> > > - PAth request contains requirements to set up a path
> > >   of *certain* protection guarantees without knowing the
> > >  topoogy and its capability
> > >
> > > Now ...
> > >
> > > If i want to set up an end-to-end *backup* path, it is the network
> > > (intermediate nodes) which has to decide if a *backup*
> > > link or a segment need to be overloaded. Don't you think in this
> > > case SE may make sense.
> >
> > I the circuit world, if you obtain a second circuit for use as backup
> > you don't need to tell the network about it.  Its just going to
> > give you a dedicated circuit, with whatever reliability guarantees
> > you have in your service contract, and with an expectation of being
> > paid for the circuit wether or not the end-system chooses to
> > put bits on it.
> 
> Not necessarily. Establishment of a secondary circuit could be a
> policy decision based on the customer SLA.
> 
> Shared mesh restoration is one case where you can see the need
> for SE based reservation for backup path.
> 
> BLSR etc where both the primary and secondary take the same
> span protected ring can be another case.
> 
> >
> >
> > In the packet world its different.  I suppose you could obtain
> > a second circuit with a service contract that allowed the network
> > to stat-mux the bandwidth with other traffic.   I still don't
> > know why you would tell the network what you wanted it for.
> 
> Agreed on this explanation.
> 
> >
> >
> > BTW. If you used such a circuit for protection it would have to be
> > 1:1, not 1+1, because the sharing on the standby channel clearly
> > means that you can't duplicate the working channel data.
> 
> Agreed. May be UPSR is a bad example.
> 
> - sudheer
> 
> >
> >
> > John
> >
> > > Thanks for your input.
> > >
> > > sudheer
> > >
> > > John Ellson wrote:
> > >
> > >
> > >>Suresh Katukam wrote:
> > >>
> > >>>Zhi,
> > >>>
> > >>>You are correct about 1+1 path protected...
> > >>>
> > >>>But if you have a LSP that is protected by some 1+1 links and some
> > >>>UPSRs, BLSRs etc.. then this LSP contains mixed protection schemes
> > >>>(I am not sure what you call this LSP - 1+1 protected, just Protected
> > >>>circuit).
> > >>>In this case, SE style can be used..
> > >>
> > >>If you're talking about nodes other than the nodes that are
> > >>at the the ends of the protection span, then I suggest that you just refer
> > >>to it as a "reliable segment".  It shouldn't matter
> > >>to the end-systems how that segment reliability is achieved.
> > >>
> > >>Protection is only interesting to nodes that have to take part in it,
> > >>otherwise its just a segment of a connection with a greater or
> > >>lesser propensity to failure.
> > >>
> > >>John Ellson
> > >
begin:vcard 
n:Vissers;Maarten
tel;cell:+31 62 061 3945
tel;fax:+31 35 687 5976
tel;home:+31 35 526 5463
tel;work:+31 35 687 4270
x-mozilla-html:FALSE
org:Optical Network Group;Lucent Technologies Nederland
version:2.1
email;internet:mvissers@lucent.com
title:Consulting Member of Technical Staff
adr;quoted-printable:;;Botterstraat 45=0D=0A=0D=0A;1271 XL Huizen;;;The Netherlands
fn:Maarten Vissers
end:vcard