When we do time series query, it is normal that it will return lots of time series data after groupBy(i.e. cpu usage group by host). And often we want to select the TopN from these series, and the sort rule is not depend a single dimension, it depends the whole value of these series(i.e. the highest 5 average of cpu usage). If we can do this, we can answer a lot valuable question, like, what’s the highest 5 CPU average usage during the past 4 hour. How many host’s incoming bandwidth run upper than 120MB/s.
Of course these can be done in front end of the engine, but we think it’s much more efficient if the query engine can support these queries.