AW: Recommended 1Gb SFP for ~115km?

["Thomas Weible" <Thomas.Weible () flexoptix net>]

Hi,

the setup with two media-converters works but has a major drawback. If you want to see the overall line (digital 
diagnostic) you always have to take into consideration that there are actually 3 physical links involved in the overall 
link. Looking from your routers you only see the SX link (basically 2 meters via patchcord). The potential trouble 
making long distance link is hidden - or you have to look on the media-converters (if management is implemented).

Cisco did a quite good job on implementing the DDM characteristics of the optics. So why not to take a 32dB or even 
41dB power budget SFP and make it workable in the switch / router. Works like charm in some setups and you see straight 
the actual line.

cheers
Thomas

PS: hello to NANOG from my site. I just got invited to the list because of this topic here. My field are the fibre 
optic networks based on pluggable technology.

-----Ursprüngliche Nachricht-----
Von: Abello, Vinny [mailto:Vinny_Abello () dell com] 
Gesendet: Mittwoch, 4. August 2010 18:27
An: Justin M. Streiner
Cc: nanog () nanog org
Betreff: RE: Recommended 1Gb SFP for ~115km?

Thanks for the input, Justin. I'm familiar with Transition Networks and have used their solutions in other scenarios 
(as well as MRV). I'm aware of the fiber characteristics being a major factor of the link budget and dispersion, etc. I 
am waiting on measurements from the company who is finishing the splicing of the fiber for us so I know what I have to 
work with.

Thanks again!

-Vinny

-----Original Message-----
From: Justin M. Streiner [mailto:streiner () cluebyfour org]
Sent: Wednesday, August 04, 2010 7:46 AM
To: Abello, Vinny
Cc: nanog () nanog org
Subject: Re: Recommended 1Gb SFP for ~115km?

On Wed, 4 Aug 2010, Abello, Vinny wrote:

> Any pointers on real world experience on this topic would greatly be 
> appreciated. What are people using successfully out there as far as 
> third party SFP's go to hit a distance of approximately 115km? This 
> would be for
a
> Catalyst 6506. Cisco's solution was a much more costly EDFA solution, 
> but
I
> see plenty of vendors that make SFP's for Gigabit Ethernet that range 
> from 115km to 150km and more. I know these are not supported by Cisco 
> and TAC won't troubleshoot if they are in the switch. I'm willing to 
> work around that should I need TAC assistance on the switch. What 
> works well for a single wavelength solution at this distance without 
> having to switch to DWDM? This circuit will have duplex fibers.

I just lit a ~110km fiber span with gigabit gear in the last few weeks, and I ended up going with an external line 
driver because the native Cisco options wouldn't work, for a variety of reasons.  I would have preferred to plug 
directly into the 6509s I have at each end, but it wasn't feasible.

I ended up going with gear from Transition Networks.  Pricing was pretty reasonable and their customer service has been 
great so far.  There are some things with the management interface that I'm not too crazy about, but nothing that was a 
show-stopper.  The driver has a 2-port SFP module that takes the LX12 long-haul signal in one side we drop it out the 
other side as 1000baseSX to drop into the 6509s.  Works like a charm.

I also looked at kit from MRV and Metrobility.  The MRV stuff looked good too.

Having said all that, you need to take into account the engineering characteristics of the fiber span, to make sure you 
choose gear that will live within the attenuation and dispersion limits that physics, fiber quality, splice quality, 
etc will impose upon you.

On the 110km span I just lit, I used the G.652 spec as a guide and figured for 0.2 dB/km of attenuation at 1550 nm and 
0.5 dB of loss for each connector, which got me an estimated loss budget of about 23 db, and the span tested out better 
than that.  With an additional cross-connect at the one end I'm still at about 22.5 dB end to end.  The LX12 optics I 
used from Transition have a link budget of 32 dB, so we have a bit of headroom.

jms