Datalad performance with dicom files

Hello. I am having issues with datalad’s performance on my lab’s cluster. I tried saving the dicom files of 75 subjects (~30k files per subject, each subject being a subdataset) but this took more than 24 hours. Using datalad status in the sourcedata superdataset also takes around 10 minutes or more. Then I tried installing the sourcedata dataset to the bids dataset (if am not mistaken, this is what YODA principles recommend doing) but this process is again too slow. Am I doing something wrong?