This message was deleted.

If your __time field still has milliseconds on it, you could try a modulo of the Epoch, e.g.:

select * from wikipedia where mod(TIME_EXTRACT(__time, 'EPOCH'),100) = 7

Your divisor value and what you compare it to will allow you to pull out virtually any random slice of records from the datasource.

Thanks for the suggestion @John Kowtko, any otherway to get a random slice from a pretty large dataset coz the above will cause the executor to go through each and every records available?

Maybe the earliest()/latest() record by time slice? something like this:

select floor(__time to MINUTE), earliest(__time,1000) from wikipedia group by 1

but use earliest to pull each column out of the group. Not sure if this is still going to fully scan the segments or go in selectively. Remember also, this is a columnar database. So it shouldn't pull the other columns until it has found which rows are satified by the filter condition ...