Comment by hbs

When it comes to time series, reasoning in terms of byte size does not really make sense, it's better to state how many datapoints you need to handle and in how many distinct time series they are distributed.