When full monitoring is needed (like for the top keys) but still the analysis must be run continuously on a production instance, maybe in the next version of MONITOR (already planned) we could say to it "give me N items and return back in normal mode", or a similar thing but with time (for instance 1 millisecond), so that it is possible to perform some kind of polling that has no serious impacts on the server. Well I'll think about this when I'll reimplement monitoring :)
In redislive, the monitoring duration is configurable via command line argument, the idea is to schedule it as a cron job and monitor every 6 hrs for 1-2 minutes, that's enough to build the trend for analytics.
