Forum Discussion
I think the numbers you are looking at are heavily damped. If you look at the daily graph (https://stats.labs.apnic.net/ipv6/AS5089?a=5089&c=GB&x=0&s=1&p=1&w=30&s=0) which seems to respond to changes much more quickly it looks like the trial is over and things are returning to their pre-trial state. With the pre-Christmas change freeze coming I doubt we'll see a deployment before the New Year.
@davefiddes writes:
I think the numbers you are looking at are heavily damped. If you look at the daily graph (https://stats.labs.apnic.net/ipv6/AS5089?a=5089&c=GB&x=0&s=1&p=1&w=30&s=0) which seems to respond to changes much more quickly
I think it's just the opposite --- that the Javascript tables of raw CSV-like data are the actual sample counts which I plot, while the graph you linked has badly messed up processing and is showing fall-off for recent days because it's averaging over the 30-day period of their default presentation, and the month hasn't completed yet. You have to use the Average Interval (Days) button to change the interval to 1 day to match the daily CSV, but when you do, instead of seeing taller and sharper peaks and troughs owing to independent samples, instead you see a broader hump suggesting that it's still using its monthly processing but resampled to give it daily resolution. That's not how you'd expect data to behave on changing its averaging window.
Although we would never be able to detect from the CSV a simple delay without other processing --- the same data being sent out as they receive in, but delayed by some days --- I see no reason why they would want to do that. And it's definitely not averaged, because we've seen near-instant drops in a day or two, without the tell-tale signs of averaging,
Can we dump data from those graphs you linked? I could tell a lot more from looking at the underlying numbers that they're plotting. I don't want percentages on the Y-axis, they're next to meaningless. In the Javascript/CSV source, we can see not only the samples but the samplecounts too --- it's much more reassuring and compelling statistically.
Related Content
- 6 months ago
- 8 months ago
- 9 months ago