[c-nsp] ONSs, DWDM SFPs, and the 3560/3750E

Michael K. Smith - Adhost mksmith at adhost.com
Mon Apr 7 16:50:43 EDT 2008


Hello Justin:

More in line below.

> -----Original Message-----
> From: Justin Shore [mailto:justin at justinshore.com]
> Sent: Monday, April 07, 2008 1:26 PM
> To: Michael K. Smith - Adhost
> Cc: Cisco-nsp
> Subject: Re: [c-nsp] ONSs, DWDM SFPs, and the 3560/3750E
> 
> Many thanks for the reply, Michael.
> 
> Michael K. Smith - Adhost wrote:
> > The filters are put in place between your 15454 and your one-pair
> uplink.  So, something like this:
> >
> > 15454 -> Lambda 1  \
> > 	-> Lambda 2   -> Filter (muxes wavelengths) -> outbound fiber
> > 	-> Lambda 3  /
> >
> > And then same in reverse.  So if you have 3 Lambdas, you will have 3
> fiber connections into the filter from 15454 on separate wavelengths
> and the output will be on one set of fibers for transport.
> 
> So, if I'm understanding correctly what you wrote and what I've been
> researching today, essentially the xponder card acts as a switch and
> uses the 10G interfaces for one of 3 L2 design scenarios (outlined in
> the line below).  No DWDM is happening yet but instead we're using the
> xponder card as a 20 port GigE switch with 10G uplinks.  Would that be
> a
> fair statement?

That is correct.
> 
> Then, if we have multiple xponder cards we could take their output and
> stuff them into muxes (15216 for example), thus introducing the
> benefits
> of DWDM.  Is that correct?  Or we could just carry the xponder 10G
> links
> around the network without DWDM and add the DWDM gear when out
> bandwidth
> approaches the 10G mark.  Am I on track or in the ditch?
> 

Yep, that would be Lambda 1 and Lambda 2 above.  You would put on card on 155x.x and the other one on 155y.y and then carry them out on a single set of fibers.

> Can the output from DWDM SFPs in regular switches be used as input
> straight into the filters?  Not that I have a use for this right now
> (unless the 3560E-12D gained support for DWDM SFPs) but it would still
> be interesting.
> 

Yep, as long as the wavelengths match.

<snip>

By the way, have you looked at the ML-series cards?  It's a different approach to the same problem, but it gives you ring failover on the back end as well.  Since you own the fibers and don't have to worry about purchasing lambdas from a provider, this may be a better solution for you from a resiliency perspective.  (IMO, of course).

Regards,

Mike
-------------- next part --------------
A non-text attachment was scrubbed...
Name: PGP.sig
Type: application/pgp-signature
Size: 475 bytes
Desc: not available
Url : https://puck.nether.net/pipermail/cisco-nsp/attachments/20080407/45df3efd/attachment-0001.bin 


More information about the cisco-nsp mailing list