Packet-Optical Technologies
Showing results for 
Search instead for 
Do you mean 

Don't Go Over that Cliff

by Juniper Employee ‎11-30-2015 03:37 PM - edited ‎11-30-2015 03:44 PM

I don't know about you, but I spent many hours (way too many hours) as a kid glued to the TV screen watching Saturday morning cartoons. Looney Tunes were always a favorite, especially The Road Runner.  If you've watched more than one episode you know it'd only be a matter of time before the hapless Wile E. Coyote would end up running off the edge of a cliff. Always funny, but he's seriously bummed.


RoadRunner Coyote.png

Why bring this up here? Well, in the packet-optical transport world we sometimes live in fear of falling off cliffs: FEC cliffs, to be specific. If the link BER exceeds the correction capability of the receiver's FEC decoder, uncorrected codewords result. Uncorrected code words -> corrupted Ethernet frames -> dropped packets -> you are bummed.
To avoid that cliff you need watch where you're going and know what to look for. Heres an example for a PTX 100G DWDM PIC:  
Step 1 - Clear the Stats
When looking at FEC stats, it's important that your readings are not stale. If you bring a link up or down, you'll likely see some residual errors that have been accumulated by the FEC PM counters. While this is completely normal and expected, you'll want to make sure you clear the counters once the link is stable to ensure you're looking at meaningful data. Here's how:

> clear interface statistics et-1/0/0

Step 2 - Read the Stats
Once you have clean stats ready, you'll want to look at some numbers. The longer your collect stats, the more meaningful they will be. Here is a CLI excerpt on how to read them (see more about the parameters here):


> show interfaces et-1/1/0 extensive




OTN FEC statistics:

    Corrected Errors                                13032402

    Uncorrected Words                                      0

    Corrected Error Ratio (        53 sec average)  3.42e-05


Step 3 - Look for the Cliff
Once you have valid stats, you should compare them against the FEC cliff for your specific interface type. Looking at Table 1 from the TechLibrary article Understanding Pre-FEC BER Monitoring and BER Thresholds, we can see that the FEC cliff for the PTX 100G DWDM PIC is 1.5E-2. 
Putting all this together for this specific example, we can see that we're in good shape: there are no uncorrected words (UCWs) and the current pre-FEC BER (aka Corrected Error Ratio) has about 3 orders of magnitude margin vs. the FEC cliff for this PIC.
Don't be that Coyote
Knowing your link performance and equipment limits are fundamental to turning up and maintaining a stable and reliable packet-optical link. Going over the FEC cliff is no laughing matter... if you do, you might wind up with an anvil on your head.

by Juniper Employee
on ‎01-19-2016 10:16 PM

I had no TV till I was 22 yrs (not months) old.. I caught wih road runner in my adult life... Smiley Happy


Juniper Networks Technical Books
About the Author
  • I started work at a router company, moved to an optical company adding MPLS, moved to mobility company, moved to a packet optical company who got bought by a router company. Full Circle!!
  • David is a Distinguished Engineer in Juniper's Optical Engineering team, having joined Juniper as part of the acquisition of WANDL Inc in January 2014. David is working on routing and optimisation software for multilayer networks to help plan and design networks using the new generation of packet-optical technology from Juniper. In "previous lives", David worked on soliton propagation; diffractive optic device design and network design software and algorithms in the Optics Research Group in BT's Adastral Park Laboratories in Ipswich. He holds a BA and PhD in Mathematics from the University of Cambridge.
  • Steven Keck is a Distinguished Engineer within Juniper's Optical Engineering team where he is responsible for architecture and implementation of packet-optical solutions for routing and switching platforms. Steve has been designing telecom hardware and optical systems for nearly 20 years. Prior to Juniper, he has held positions at Stanford Telecom, StrataLight Communications and Cisco Systems. He holds a B.S. in Computer Engineering from University of the Pacific.
  • David Song is a Sr Staff Engineer within Juniper's Optical Engineering team where he is responsible for the design of packet-optical solutions for routing and switching platforms. He joined Juniper in 2004 and has been designing networking software in control plane and data plane on various platforms. Prior to Juniper, he held various software development positions at Ciena and Nortel Networks.