Everyone wants a reliable network - it's good for users and good for your long-term employment. Unfortunately, $&!* happens - and packet-optical networks are no exception. No optical link is bulletproof, but one can improve the reliability and resiliency to packet loss by ensuring adequate system margin.
What is system margin? Think of it as an insurance policy for your optical link. Car insurance helps protect you from financial ruine in the event someone hits you on the road. Similarly, system margin helps ensure your packet connectivity isn't ruined if some unexpected bit errors hit your optical link.
Mind your Bs & Qs
In previous posts have spent time talking about FEC and BER. These two concepts are fundamental to understanding packet-optical link performance. That in mind, when evaluatin performance is would seem intuitive to just compare the difference between two BER values, right? Sort of...
Let's take a look at two examples where two BER values differ by one order of magnitude:
A) BER 1E-8 vs. 1E-7
B) BER 1E-3 vs. 1E-2
It seems as if the difference is both cases is the same, but it turns out to not be the case. The challenge here is that BER values are logarithmic in nature and that can make evaluation difficult (especially for this author) since one has to compare non-linear values. If you don't do this often, the raw numbers can be misleading.
To make our lives easier, let's do a little math. To compare performance metrics BER values can be converted into a representation called 'Q2-factor', colloquially known as 'Q'. To see how to convert BER to/from Q (and vice versa) check out this TechNet article.
Let's revisit the earlier example using BER and Q:
A) BER of 1E-8 = 15.0dBQ, 1E-7 = 14.3 dBQ -> a difference of 0.7dB
B) BER of 1E-3 = 9.8dBQ, 1E-2 = 7.3 dBQ -> a difference of 2.5dB
As you can see, the looking at the performance delta in dBQ terms paints a very different story vs. the raw BER values. If you recall, the prime directive in optical transport is to avoid falling off the FEC cliff. If we were to pretend that the latter value in each example scenario was the FEC limit, you can see that there would be a dramatic difference in the safety margin in the two scenarios. Suffice it to say it's vital to look at BER numbers in the proper context.
Now that we can look at link performance in terms of Q we can revisit the original system margin question. Let's say you've successfully turned up a PTX 2x100G DWDM PIC and the link is up. We're ready to figure out the link margin:
Read the BER stats from the CLI: For this example, we'll assume a hypothetical value of 3.4E-5
Convert your BER value to Q: Using the earlier reference and a little Excel we find that 3.4E-5 = 12.0 dBQ
Know your FEC limit: Looking at Table 1 from the same article we see our PIC has a FEC limit of 6.7 dBQ
Calculate the system margin: Take your measured Q value and subtract the FEC limit, 12.0 dBQ - 6.7 dBQ = 5.3dB
There you have it: Your link has 5.3dB of Q margin - not bad (assuming this isn't a loopback cable!). To consider this in in optical terms, this would be comparable to (but not quite the same) as having 5.3dB of ONSR margin.
So now you have a means to calculate system margin, but how much is enough? That's a very subjective question that ultimately boils down to how aggressive or risk-adverse you want to be.
If you want to eek out the maximum reach possible (e.g. to avoid regeneration and reduce capex) and your network can tolerate a rare hiccup once in a great while, then perhaps 1dB of margin would suffice.
On the other hand, if you need maximum reliability above all else you're going to need more margin. 2dB or even more may be appropriate
Back to our earlier example: If we had only 0.7dBQ system margin it woudl be fair to say we were skating on thin ice. A slight hiccup anywhere on the line system or endpoints could push you over the FEC cliff and result in packet loss. By contrast, 2.5dBQ would be considered comfortably safe and would offer a significantly greater safety margin.
Regardless of your risk tolerance, it's important to have a system margin target for all of your packet-optical links. With a target in mind and the know-how to measure it, you can increase your confidence in your long-term link performance. Without knowing your system margin, you may be rolling the dice... and you know what might be just around the corner.