Hi,
Targeting to use smart baseline alarming on a percentage of modems offline and seeking to avoid alarms being fired for segments that might see a significant relative increase in the number of modems offline, while the figure itself is still very low (e.g. increase from 2% to 6%), I would appreciate to learn if possible to add a rule so alarms would only fire if the percentage itself would exceed a given threshold.
Thx
Not sure if this is a meaningful comment in the context of your question, but just in case it is: for metrics that indicate % online/offline users, I have seen users typically go for the option to calculate the baseline value as the mean of the values at the same time of the days before, for x days. So you would then compare the % online users at 9AM today, with the mean value of the % online users at 9AM yesterday, 9AM the day before that, etc.
But you seem to indicate that you want to detect only big deviations? Could you not just do that by defining an absolute variation on your %-metric? E.g. only alarm when there is a deviation of more than 10 points on your percentage value (it's a bit confusing to talk about it, because your metric is a percentage, and on that percentage you can do an absolute and relative variation).
Or did I miss the point of the question here?
Great, glad to hear it worked out.
HI Ben,
Thank you very much for your feedback here. It did work out indeed just to switch from a relative to absolute alarm thresholds.