Is there value in segregating rare tag combinations



Hi All,

i am monitoring a mid sized network (600 devices or so) with telegraf/influx, and i was curious about if it is better to store a similar metric (interface counters) all in the same bucket, or to split similar tag combinations into metrics…

Example, only 10 of my devices have interfaces starting with “Ethernet” (for the “ifDescr” tag) and share next to no common ifDescr’s with the rest of the devices dumping into the interfaces counters metric…

the advantage of this seems to be that i can run the same CQ over this data, but i did find after i pruned some of the more useless counters that my CQ runtime dropped significantly (30 seconds to 17 seconds).