That omits most interesting part: the fundamental data model. For example, Prometheus is "stat collector": it polls the inputs at pre-defined interval, and onl

That omits most interesting part: the fundamental data model. For example, Prometheus is "stat collector": it polls the inputs at pre-defined interval, and only record floats. In Prometheus, you cannot record "failover event" directly, the best you can do is "number of failover events in last X minutes". At the same time, InfluxDB is "event database", which accepts any data (floats, strings) at any intervals. This is nice as you can, for example, plot "failover event" at the same graph as "system latency" and see how failover affected latency. But as a result, the queries need to be more complex to account for possibly-irregular data. _This_ kind of comparison is very important, and can eliminate products early.