My co worker used our local deployment of the geneva forecasting engine to see how fast we could forecast a million time series. With parallel processing we hit 2300 forecasts per second on only 32 CPUs of compute - a million forecasts in 7 minutes 9 seconds. More stats in article.