Optical / DWDM

Started by javentre, January 07, 2015, 06:46:15 AM

Previous topic - Next topic

javentre

Just check to see how many of you have experience in transport / optical (DWDM) networks?
[url="http://networking.ventrefamily.com"]http://networking.ventrefamily.com[/url]

Network2501

What kind.... I can talk colours/channels all day. But generally deal with passive gear.
- Pete

javentre

My primary focus is on Cisco MSTP (active).
[url="http://networking.ventrefamily.com"]http://networking.ventrefamily.com[/url]

killabee

We're just delving into DWDM now, specifically looking to get multiple 10G waves (8, 16, 24, or 48, I believe) out of our limited dark fiber.

I'm not the decision maker or the one purchasing the equipment, but hopefully this interests us enough to get more.  And hopefully within a few months I'll know more about it.

Network2501

Have you read much about it? My first introduction was a presentation at Nanog which covers most of the basics.

Tutorial: Optical Networking 101 & 201
https://www.nanog.org/meetings/nanog48/agenda
- Pete

killabee

Actually, thanks for reminding me about that NANOG video.  I've watched it before but will need to watch it again now that it's becoming a reality.  I've only done very high-level reading, such as CWDM vs DWDM, what they do, and the like. 

Network2501

For the most part it's pretty simple. In my humble experience it becomes more complex is when you're trying to push optics to their limits or provide services via active systems. Which really is just a question of how you provide/manage services.
- Pete

routerdork

I dealt with a few different vendors at the ISP I left last summer. The main product was the Cisco MSTP M6. I can't remember the name of the design but I was spending a lot of time on that. CTP I think it was.
"The thing about quotes on the internet is that you cannot confirm their validity." -Abraham Lincoln

javentre

Yes, CTP is the design tool.

My biggest complaint against Cisco's DWDM stuff design/monitoring tools is that they're Java based.  I have a strong dislike for java.
[url="http://networking.ventrefamily.com"]http://networking.ventrefamily.com[/url]

killabee

Looks like we'll be getting the ONS 15216 (odd channels) and the a few of the odd channeled optics.

Based on my reading these passive muxes are pretty straight-forward.  You install an optic into your device that corresponds to one of the usable channels on the mux, cable it up to the corresponding channel on the mux, connect the fiber span to the composite ports on the mux and demux side, then on the farend demux cable the same channel to your device and use the matching channel optic.  Something like this:

Switch [Channel1-Optic] ---> [Channel1]MUX[Composite] -------> [Composite]DEMUX[Channel1] ---> [Chanenl1-Optic]Switch

Does that sound right?

routerdork

Quote from: killabee on February 02, 2015, 12:05:23 PM
Looks like we'll be getting the ONS 15216 (odd channels) and the a few of the odd channeled optics.

Based on my reading these passive muxes are pretty straight-forward.  You install an optic into your device that corresponds to one of the usable channels on the mux, cable it up to the corresponding channel on the mux, connect the fiber span to the composite ports on the mux and demux side, then on the farend demux cable the same channel to your device and use the matching channel optic.  Something like this:

Switch [Channel1-Optic] ---> [Channel1]MUX[Composite] -------> [Composite]DEMUX[Channel1] ---> [Chanenl1-Optic]Switch

Does that sound right?
Yes it does. I did this with ASR9K's to cut down on the fiber between buildings for nV connections. As long as the optics are supported by the switches you should be good. Make sure you check those light levels as well.
"The thing about quotes on the internet is that you cannot confirm their validity." -Abraham Lincoln

routerdork

Quote from: javentre on February 01, 2015, 07:22:14 PM
Yes, CTP is the design tool.

My biggest complaint against Cisco's DWDM stuff design/monitoring tools is that they're Java based.  I have a strong dislike for java.
That was the biggest issue we had with it as well. Compared to some of the other DWDM products we had used Cisco was pretty solid though.
"The thing about quotes on the internet is that you cannot confirm their validity." -Abraham Lincoln

javentre

I've had far fewer issues with the optical gear than any other product from Cisco, by a wide margin.
[url="http://networking.ventrefamily.com"]http://networking.ventrefamily.com[/url]

killabee

We got the mux and optics installed and our links successfully came up, but the light power readings got me a bit concerned.

Initially, I tested the fiber span end-to-end using ER optics on both sides, and the Rx receive power was around -9.66 to -9.8 dBm (the Tx power on the opposite side was around 0.7 to 0.82 dBm).  After we installed the two muxes and patched the extra fiber, the Rx power dropped between -16 to -20 dBm!  Based on "show interface X transceiver details" the interface won't warn or alarm until the receive power drops to -24 dBm, so the links are still up, but if we add any more patches, disconnect/reconnect the fiber too many times, or splice any more then we'll incur the extra loss needed to take the links down.

Could the mux be adding that much loss?  I'm having our fiber guys clean all the patches just in case, but if they were dirty in the first place then I would have seen the loss during my test with the ER optics.

javentre

#14
Yes, they can take that big of a hit on the optical budget.  Which model did you buy?

The 15216-MD-40 has an insertion loss of 4.5-6.5 dB, but you also need to account for connector/patch loss too.  Coupler/Connector loss is generally .25-.50dB

Fusion splice where possible, avoid needless patching, and clean your connectors.  It's not like 1GE and 10GE in LAN environments, it does matter.
[url="http://networking.ventrefamily.com"]http://networking.ventrefamily.com[/url]