@Nuno Nogueira From the screen shot it looks like your experiment is running with some modules run queued. Depending on the compute resources you have chosen the run time might vary and with a free trial account a lower compute power will slow down the run time. I have tried the same experiment with a STANDARD_DS2_V2 type with scaling option from 0-4 resources in the cluster and the run completed in around 23 minutes where most the time went to setup the compute since I do not have any running compute to run the experiment directly. You can try a similar setup with a higher compute and check if it runs faster.