On Thu, 13 May 2021, Gunnar Wolf wrote: Dear Gunnar,
> > Okay, but it would be useful if you could standardize the graphs somehow. > > I.e. try to PLOT successful_connections per time_interval. > > I don't know if I understood this correctly. As I said, I changed the > polling frequency after a couple of months; I am sure this would skew > the results. I mistakenly thought the big drop on the graph in 2019 may be derived from changing sampling method: | Do note that at the beginning I sampled the network much more often | (hourly). I decided this was too much, and since June 2019, am now | walking the network only every four hours. I do hope nobody sees this | as excessive! | This shows the very large drop the SKS network had in mid-2019, as | well as its behavior since then. I am happy, even hopeful, to note | that it seems the network hit reliability minimums between October | 2020 and February 2021, but it seems there is a slight trend for | improvement, at least back to the late-2019 levels. So I wanted to suggest two possible solutions: * Simply delete intermediate data and keep just every fourth sample in the dense sampling era. * Plot the ratio of the different results in percent instead of the absolute counts. > My first plots had one column for each sample. What I'm doing now > -precisely, in order to normalize the plot- is to "lump together" all > of the plots for a given day into the same column. Thus, the generated > plot respects and properly represents time (skipping a day here or > there, as there might be some glitches in my data acquisition), but > mostly in an uniform way. > > (...or... Did I completely fail to understand you?) No, it was me who misunderstood the graph. I did not see by naked eyes that samples are denser on the left side of the horizontal axis. I thought there is one sample per day and you simply added the whole day numbers before plotting. Cheers Gabor
