perfSONAR User Q&A and Other Discussion

[Mark Feit <>]

Michael Reece writes:

I also understand that the latency isn't extremely accurate (perhaps not accurate at all?) and should be taken with grain of salt, but was told to have latency as one of metrics in the MaDDash. If the latency value is truly worthless at this range, maybe I will speak to my boss about removing it entirely.

Inaccurate as it is at short range, there may still be value in measuring latency if you’re not too concerned with the absolute figure.  Between the laws of physics and the accuracy of your timekeeping, it’s possible come up with a rough number for the upper end of “normal” behavior for segments on your network with the inaccuracies factored in.  (You should be able to correlate that with what you’ve observed previously.)  Measurements below that number may not be accurate, but they do tell you that things are good within some limit of believability.  As the actual latency increases, the influence of the timekeeping error becomes smaller, but crossing the line may be a good indicator that things may be headed off the rails.

I was more interested about the changing the color of the latency grid on MaDDash since the red makes it seem like there is major problem. Mark, I like your idea about changing the idea the result to an absolute value. Since the maddash.yaml doesn't contain the code where this occurs, where exactly do I make this change? Just not sure where the file would be located.

MaDDash isn’t my area of specialty, but I think Ivan’s comments should get you moving in the right direction.

—Mark