I'm in the middle of a long running project to choose the best monitoring toolset we can. Right now I'm testing ELKStack and Prometheus+Grafana. During my research I remember reading about the idea of merging old metric data into fewer data points.
So, if you have data from every 1m, you would merge 15m worth of data into one data point.
The basic idea is that you don't need a very high resolution on data more than N days old. So you can reduce your long term storage by 15x.
Since storage is one of the resources my workplace is low on, the idea took root.
A) Is it a good idea?
B) Has anyone implemented it?
B.2) If you've implemented it, how has it worked out?
B.3) If you've implemented it, how did you do so?